/sci/ - Science & Math

>>9206366
>Does race exist as a biological concept?
It does but you're gonna fail your class and be ostracized if you say so.

>>9206366
Race does exist but not in the broad sense of "white", "black", "asian", etc. Different types of humans are as specific as different types of other animals. Just as there are dozens to hundreds of specific dog subspecies, there are also dozens to hundreds of specific human subspecies. Like for example, not all black people are highly susceptible to sickle cell, the susceptibility varies widely depending on whether the black person's ancestry is mainly from North Africa, Central Africa, the Caribbean, etc.

I think it's undeniable that humans have their own subsets of race, breeds, species, whatever. All other animal species have it, and we aren't special or above other animals. The real question should be whether or not it actually matters, since unlike dogs, humans are self conscious advanced creatures that are capable of being more than their instincts and genetics. But that would be getting into /pol/ discussion.

>>9206366
Lots of different sub-species who all evolved to suit their environment. None is superior to the other as the only pressure is survival, which all existing ones have succeeded in. They do have different traits though and you could argue that combining them or placing different groups in the other's environment is deleterious.

>>9206366
Sickle-cell anemia exists and it disproportionately affects the "brown people".

I think that "brown people" also experience higher rates of vitamin D deficiency when living far from the equator.

White people get sunburn.

If you just look at the medical literature, I think you will find that certain races are disproportionately predisposed to certain afflictions.

In the medical field, race certainly does exist if you want to call it race.

>>9206366
Hispanics are caucasoid!
Anyway it exists biologically and other posts have said this, but honestly just do the prompt and tell them what they want to hear

>>9206366
>We have to do a assignment in Physical Anthropology explaining why race doesn't exist biologically

No, you don't.

>>>/pol/

>>9206537
He specifically said he didn't want to go into Paul territory

>>9206366
>We have to do a assignment in Physical Anthropology explaining why race doesn't exist biologically
That is weird.

In science you set out to explore a topic and then draw a conclusion as far as you can do that which never is given. You do not go in to a field to prove something. You can however falsify a theory or hypothesis but that is, strictly speaking, not quite the same thing. It might sound like splitting hairs but if you want to follow the scientific method you have to frame the question very carefully.

>>9206504
>It does but you're gonna fail your class and be ostracized if you say so.
True.

You can get around that by saying you instead have ethnicities which somehow is accepted.

>>9206523
>evolved to suit their environment
One example is the shape of the nose: long in dry climates (like Scandinavia) and short in humid climates (like Japan).

>>9206526
Ability to metabolise milk proteins differs greatly between ethnic groups, same with alcohol. Forensic science has a lot of means to determine ethnicity. Look up ear wax colour as one example.

>>9206366

>Physical Anthropology

>Learning a meme (humanities) science

Kek, I believe that it is unanimously accepted as a concept in medicine (if it wasn't such would result in the illness or death of the patient).

I suppose you could try cherrypicking data and use strict definitions of subspecies and so on to make the case against 'races'.

This would allow you to both present the case that significant and meaningful genetic variance exists, while also not appearing like a '/pol/tard' or Nazi, etc.

>>9206366
No, they don't.

>Genetic differences between humans do exist.
>There are many different techniques for attempting to measure "genetic distance" between humans.
>Given different groups of humans there are techniques for measuring genetic distance both within the groups and between the groups. This all falls under the theory of population genetics.

Unfortunately:
>There is no one-to-one mapping between a gene/mutation and a physical characteristic.
>Genes produce proteins, so mutations in genes can lead to widespread effects throughout the body, though most are usually imperceptible (eg. mutation in gene that produces cell walls can lead to deafness and other shit).
>Given a physical characteristic there are tons of ways to achieve it via genetics and due to convergent evolution it isn't uncommon for this to happen. See pic, left/right groups are genetically similar but physically similar fish are genetically distant.

Moreover:
>Epigenetic data and environment both control when/if genes activate. So, physical characteristics are dependent on environment.
>Differences themselves may be geared toward certain environments. Corn that does well in sunny dry regions vs corn that does well in cloudy rainy regions. Put them both in a sunny environment and one will seem "better" than the other, but really it's just the environment.
>Deriving social characteristics from genetics is incredibly hard. The best modern techniques attempt to use genome wide polygenic scores to guess stuff like edu-years but even then they're still in infancy.

Race is a pre-genetics approach to biology. They basically would take physical, cultural, linguistic, etc.. characteristics and try to create broad classifications to capture them (primarily physical). Unfortunately it's a shit tier classification and the approach has been entirely rendered obsolete through population genetics.

(cont.)

>>9206619
(cont.)

All of that said, race is so ingrained in society that the social sciences REFUSE to let it go claiming it's useful to their research. As such race has been abandoned by the hard sciences but the social sciences have salvaged it as a "social construct". In that context race is typically studied by giving people a questionnaire that asks them "what race they wish to identify as" (they do not ask them about ancestry or perform any sort of biological test because such approaches have been proven garbage by the hard sciences). If you read a news article about race then you should check the paper's it cites because more than likely it's either a paper on population genetics that the journalist added a bunch of their own assertions to or a social science paper (or just some oldschool pre-genetics paper based on assertions that are now known to be wrong).

If you want to make robust assertions on a personal level then you should focus on the individual.
>Is this person prone to violence
>Does this person carry the gene in question
>etc...

Regarding genetic testing: Privacy laws in most countries aren't caught up to this. It could be possible for an employer/landlord/government/etc.. to ask you to provide such records if they exist (and if one does then you should immediately contact your country's privacy commissioner if you have one or something like the EFF or ACLU). For now it's maybe best to avoid such tests until appropriate privacy laws are established.

>>9206526
But these afflictions are genetically mediated... far more complex than the racial labels used. Thats the point race labels we use are unscientific evenhough genetics obviously varies continuosly across the world in a sense that reflects ancestry. Examples of this complexity are how for instance certain european populations have higher instances of sickle cell anemia than certain african ones. Theres always exceptions and its mottled.

Race is one of the most garbage, pseudoscientific concepts I have ever seen. The US keeps pushing it for some reason, and it's really pissing me off at this point. I live in a country were most black people are at least 20% genetically white, yet they are put in the same classification as any African that doesn't have a single drop of European blood. If we're going to vainly classify human beings by their ancestry, let us at least use a more accurate and complex system that is actually worthy to be used in science.

>>9206366
Race doesn't exist.

>>9206366

Race doesn't exist unless its white male privilege, which exists objectively.

>>9206842
>The US keeps pushing it for some reason
As race theories go the US one is really weird.

Hispanic: different from Caucasian? How? Last I heard they came from Europe.

Jews: apparently Caucasian. Yet Arabs are Asian. While both belong to the Semitic group of peoples and both consider Abraham as their common patriarch.

What is the thinking behind this mess?

>>9206519
>All other animal species have it
but they don't all

>>9206586
>long in dry climates (like Scandinavia) and short in humid climates (like Japan).

Are you retarded? Scandinavia is far from dry

>>9206977
>Scandinavia is far from dry
West coast with cities like Bergen basically lives in rain. Other than that it is pretty dry. Winters with -30 C also tend to be pretty dry. Sure, we are not talking Sahara here but it is still dry.

>>9206526
>disproportionately predisposed to certain afflictions
Could that not just as well be because of general lifestyle differences?

>>9207045
Some yes, but some are genetic diseases. To have a disease be more prevalent in one race than another is meaningless though. All sorts of subpopulations are plagued by various genetic diseases.

The unique fact about race is that Amerindians are superior to europeans. Their development rate is superior to europeans'.

>>9206526
Brown is an extremely wide category and sickle cell anemia is largely absent from everyone outside of Africa

>>9208354
Read OP dumbass.
>>>/pol/

>>9206366
Define 'race'
Do you consider someone with a 7 inch dick a different race?

>>9206593
I abandoned anthropology out of a nascent sense that I was learning insular, politically motivated nonsense. Graduated as a Chinese major, read humanities on own time, starting STEM next semester. Best (at the time self-destructive) decision of my life.

>>9206900
IQ is also an invalid measure of intelligence and all human groups would have the same IQ with the same opportunities, but Ashekanzi Jews also have the highest IQs and it explains their success.

Define "exists"
It exists as a pattern in genetic distributions in the population, and as the physical traits derived from those genetics

It does not "exist" in the sense that it's a hard-coded, inherent part of humanity. Humanity could easily exist without any clear racial distinction, but evolutionary mechanisms have created some.

>>9206366
It doesnt matter.

The humans from Africa with the simian features seem to be prone to criminality.

>>9206903
Caucasoid not Caucasian retard.

Caucasoids are the West Eurasian human or the whitey looking face group. All humans who have faces close to whites are Caucasoids originating from a common ancestor.

>>9206366
wouldn't be the correct term be ethnicity?

It depends on what level of detail you're looking at and what you mean by race. It's a useful classification in medicine because a Swedish dude probably doesn't have sickle cell anemia, but it's artificial in the same way species is, it's a dumb assignment because you can't argue if a definition is real or not, as they're always somewhat abstract

>>9208562
>Caucasoid not Caucasian
Do you know what the -oid suffix means?

And how does this even remotely explain the Semitic part?

>is race real
>pls be /pol/ free

Every fucking time lol, inb4 300 replies.

>>9208535
>Implying the motivation and focus don´t change

>>9206366
Yes it does, bones structures are different between European and African, some studies say that some teeth diseases happens exactly because of that, two different bones structure joining in.

sub-Saharan blacks are exclusively affected by falciforme anemia(don´t know and do not want to search the name in english).

It is like comparing two dogs, one is a German Sheppard the other is a pitbul, of course a pitbul could protect your child if trained properly, but you will be better with a German Sheppard as they will be less likely to be violent for no fucking reason

>Also keep this thread /pol/ free, the image was the only picture I had related to race.

>>9209418
>you will be better with a German Sheppard as they will be less likely to be violent for no fucking reason
That's complete bs

>>9206903
>Hispanic: different from Caucasian? How? Last I heard they came from Europe.

You obviously know nothing of the history of the Hispanic people.

>>9209418
Neither of them are violent "for no fucking reason" you retard

>>9208551
I doubt it.
http://thealternativehypothesis.org/index.php/2016/04/15/ethnic-diversity-and-social-cohesion/

>>9209855
>thealternativehypothesis

>>9209886
It's a bit different than the other braindead articles.
Just give it a chance and read it.

>>9209942
They are experts in cherrypicking the tiny out-of-context bits from all these different (often outdated) studies, to form their own narrative. It should be obvious to anyone that can analyze information.

>>9209955
I know but this time they are willing to show exceptions and counter arguements

>>9209970
They always do that. All part of the narrative.

sub-saharan africans have zero neanderthal DNA unlike the rest of the world. is this not clear evidence that there are significant differences among human populations?

>>9210015
Cite your sources.
Also varg is not a source.

>>9206366
Who made this fucking picture? Jews are caucasians, also Indo-Aryans (Indians, pakis, Iranians...)

>>9210454
/pol/

>>9210454
>non-whites are white
>>>/bol/

>>9209579
it isn't you moron

>>9209731
read the news, semen slurping sperg

>>9210345
there were no neanderthals in africa you dumbfuckingass

race does exist as a biological concept
or maybe not a biological concept
maybe one step lower on the "career distribution" chain.
Like the only people that are interested in concept of "Race" and "sub species" or whatever, would be people that are invovled in animal husbandry.

My conjecture is that race does exist as a biological concept, or maybe as a physical concept. Its just that humans have used technology to gain evolutionary advantage to where race isn't really a trait selected for adaptability.
We don't need darker melanin receptors in our skin cells becuase we have been using sun block for the past 100 years.
We don't need white skin and straight hair because we have warm clothing and hair straighteners.

>>9210544
so maybe not as a biolgical concept, or a genetic one, but an evolutionary concept, which is what most of biology actually deals with when doing it's job.

I mean the traits that people group together to signify race, is often selected for when it comes to inheretence.so it does play an significant role in mate selection, and "genetic drift".

>>9208382
I just made an observation how Amerindians are superior to europeans, how is this hard to get?

>>9210703
>amerindians are superior to europeans
but who conquered who?

>>9210723
>tfw too superior to fend off enemies

>>9210738
if that is your definition of superior, then all of africa is superior

>>9210723
Natives beat the Incas. eurangutans just kept backstabbing everyone even their own kind.

Inca superiority is demonstrated by their higher development rate. It's pretty simple. Who is superior? A genius or a monkey with a gun? The answer resembles the same comparison I made about Amerindians. They developed faster. How is this hard to get?

>>9210761
yes its a true inca very good very stronk

Race is a cultural construct. Ethnicity is a biological concept. It's like gender and sex.

>>9210765
So you agree with Amerindian superiority? Great for you.

> We have to do a assignment in Physical Anthropology explaining why race doesn't exist

I didn't know you went to Good Goy State, anon.

>>9210780
>Also keep this thread /pol/ free
Bad Goy

>>9206366
Of course it does. You can send a dna sample and they will tell you what races there are in you. Unlike lets say the faggotry, which is a choice and a social construct, or other made up genders besides male, female and anomaly. What you can find out with a dna test is definitely biological.

Someone raised the genetic question using AMOVA

>>143646605
Someone raised the genetic question using AMOVA

Someone raised the genetic question using AMOVA

>143646605

Someone brought this up from a genetic point of view in an archived thread.

"I think the AMOVA used in this study is incorrect. Everyone here knows there are only 3 races aka subspecies (Caucasoid, Negroid, Mongoloid). I have underlined where they input 5. The effect of this is that races (subspecies) in humans do not exist bc the fst % according to their calculation does not surpass the 25% threshold hold applied to chimpanzees to prove subspecies exist in them.
In addition im not sure how they got 52 for population. It would be great if someone could answer that.
My question is can weaponized autism answer that races in humans do exist according to their standard if we change that number to 3 and possibly a more appropriate “number of populations” # in order to reach the magic 25% fst Threshold?
Link provided below to study
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3737365/
"

Yes, medics have to take it into account when administrating some medicines. There is also the morphology of different people's. Manny times race can act more like a trend than a Law, but it is not to be ignored

I know that a pink world where everyone is the same is a pretty idea, but groups of people who developed in different environments and different neighbours will be different

>t. Medfag

>>9206366
Y'all some racist mother fuckers. Pay up whitie

>>9211133
>Two holes above and to the side of the nose of Wast Asian Mongoloid male.
What are those? Typical features or just a random deviation?

>>9211392
they all have them, the 2 hole just above the eyes however only appear in caucasians and east asians, incidentally the races with the highest IQs

coincidence?

>>9211397
>they all have them, the 2 hole just above the eyes however only appear in caucasians and east asians
Do these have a name and a function?
>coincidence?
You tell me.

>>9211435
https://en.wikipedia.org/wiki/Supraorbital_foramen
The other ones probably have it as a notch instead where the nerve passes. Might have something to do with skull shape, definitely.

>>9211466
>https://en.wikipedia.org/wiki/Supraorbital_foramen
Thanks! Interesting what people here know.

The article was a bit confused and mixes foramen with notch without making things clear. The Talk page is also confusing. Variations are mentioned but they sure skirted that issue.

Any skull samples from our Semitic friends for comparison?

New anon here, but it depends on the frame of reference for deciding what you might call genetic differentiation. In the case most commonly looked at, it's the ratio of the genetic variation within a population to the genetic variation between populations that determines subspecies. The reason this is the most common measure is because it's the only measure that accurately controls for the specific breeding patterns/generation time of a species, and merely looks at how deviated a population may have become holistically via selection.

Think of it like this. You have a species with an incredibly long generation time, like elephants, and you try to compare that with, say, ants. Among the ants, due to the fact that there are many generations, and large numbers within populations, there is going to be an incredible diversity of genes, as there's a possibility for any one of the offspring (with many offspring occurring in a short period of time) to be born with a mutation. Thus, there's a greater genetic diversity within the population, compared to elephants which will be relatively homogeneous, with a lot more evolutionary dependence on physical events that split populations, bring them together, and the sort for there to be any kind of genetic diversity/new alleles in the gene pool.

It'd be pretty retarded to then say that because ant #132529 and ant #134000, which have a greater total # of genetic differences from each other, yet live in the same anthill as a part of the same population are separate subspecies, on the basis that Asian and African elephants are considered separate subspecies though they have less total variation.

Science is so beautiful, yet so tricky at the same time. On one hand, it is a truly amazing tool for analysis of the world. On the other, if you look at it on a surface level, without understanding, you can misinterpret so much, and come out of it more retarded than if you'd never encountered. Poor thing.

>>9211484
To add on to this post:

For your education, at your leisure;
http://science.sciencemag.org/content/298/5602/2381

http://www.nature.com/ng/journal/v36/n11s/full/ng1435.html

http://www.sciencedirect.com/science/article/pii/S0960982205002095

http://science.sciencemag.org/content/351/6273/564.full

https://www.ncbi.nlm.nih.gov/pubmed/24032721

https://www.ncbi.nlm.nih.gov/pubmed/14655871

http://www.nature.com/nature/journal/v523/n7561/abs/nature14618.html?foxtrotcallback=true

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3514343/

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC403703/

http://www.genetics.org/content/176/1/351.full

/pol/tards and non-sciencefags might not get this, but there's no quantifiable genetic way that isn't arbitrary as fuck to split humans into racial groupings. The genetic variation between populations of humans is LESS than within, meaning that two European white people on average would be more genetically dissimilar than a white person and an African.

And if the argument becomes "b-but we can have a counsel cherrypick phenotypes and assign weights to them based on some arbitrary perception of societal benefit", fuck outta here bud, you're becoming the pseudoscientific sjw that believes in more than 2 genders based purely on perception of the world and one's own desires clouding their judgment.

>>9206366
Ask them how https://www.23andme.com/ can know everything about your ancestry from your spit if race isn't a biological concept.

>>9206366
>just throw in some pictures of skulls
>And faces
whoops, how did we miss that?

>>9211495
>he genetic variation between populations of humans is LESS than within
Interesting. And how is this possible?

>>9206619
Is population genetics a fun thing to study?

>>9211832
Not by a long shot. I literally only went into it to prove /pol/ wrong.

>>9211868
>I went into a field of study because idiots on the internet are wrong

>>9206366
I think the word you are looking for is ECOTYPE. Shorthand for strain. There are different strains of human races since race as a subspecies is proven wrong by AMOVA >>9211098

See link below.

https://en.wikipedia.org/wiki/Ecotype

PS I'd really like to know what you get on this paper and what path u decide to go down lol

>>9208354
Short bus tier.

>>9212228
You mean, superior beings tier.

>>9212605
Whitey pls.

>>9212605
>.gif

>>9210761
Why did you migrate to /sci/? Are you banned from /his/

>>9213023
I just made an observation about Inca superiority. How is this hard to get?

>>9213023
no one takes him seriously enough there.

>>9213120
just because stormucks insult him with buzzwords it doesn't make it invalid what he says

>>9206619
>>9206634
>thoughtful, well formulated, evidence-backed scientific discourse

>>9209418
>sub-Saharan blacks are exclusively affected by falciforme anemia
That isn't true.

>>9213023
How would they ban him though? He just speaks the truth.

>>9213193
lel he practically is a buzzword.

>>9211495
>The genetic variation between populations of humans is LESS than within, meaning that two European white people on average would be more genetically dissimilar than a white person and an African.

This legitimately smells of pure bullshit. Im not a geneticist nor do I have a bio background but how can this be? They must be excluding parts of the chromosome that arent compatible or performing the study in such a way as to further their agenda.

>>9214146
Is it even mathematically possible for two sets to have more pairwise variance within them than between?

>>9214146
Here's the article on the matter:
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1893020/

Abstract:
>The proportion of human genetic variation due to differences between populations is modest, and individuals from different populations can be genetically more similar than individuals from the same population. Yet sufficient genetic data can permit accurate classification of individuals into populations. Both findings can be obtained from the same data set, using the same number of polymorphic loci. This article explains why. Our analysis focuses on the frequency, ω, with which a pair of random individuals from two different populations is genetically more similar than a pair of individuals randomly selected from any single population. We compare ω to the error rates of several classification methods, using data sets that vary in number of loci, average allele frequency, populations sampled, and polymorphism ascertainment strategy. We demonstrate that classification methods achieve higher discriminatory power than ω because of their use of aggregate properties of populations. The number of loci analyzed is the most critical variable: with 100 polymorphisms, accurate classification is possible, but ω remains sizable, even when using populations as distinct as sub-Saharan Africans and Europeans. Phenotypes controlled by a dozen or fewer loci can therefore be expected to show substantial overlap between human populations. This provides empirical justification for caution when using population labels in biomedical settings, with broad implications for personalized medicine, pharmacogenetics, and the meaning of race.

>>9206366
do you think that different dog breeds are the same?
they are all similar in many regards, but clearly have differences.
human races are all more similar than the the large variation found in something like doggos. the physical differences between human races are therefore less noticeable.
they exist though.
it is kinda like how blonde people are telepathic and how blacks are more likely to commit felonies.

>>9214492
That article does not say that variation between populations is less than within.

>>9211495
>The genetic variation between populations of humans is LESS than within

Why the fuck is Lewontin's fallacy STILL being posted in 2017? This fallacy only works by taking advantage of the ignorance of people who don't know how genes work. Or they'll realize how fucking stupid it is. How its an idea promoted by a marxist that never actually influenced population geneticist. How the same fucking concept can be applied to genders. look guys, genetic variation between genders is LESS then between population. Its only one measly chromosome, what difference does that make? Men and women look, behave, and act exactly the same. They'll realize that the only reason race deniers say this is so they can push their agenda.

Watch the Alternative Hypothesis. Learn it. Then stop denying race.

Look at the people denying race. I haven't read the thread, but I bet that IQ denying negro is here. He's the psychotic negro in denial who thinks Incas are better then Europeans because development, but this is a dumb idea since if that were true, then Incas should have quickly adopted to European way of life and turned into a world power ala Japan. They didn't, because their IQ is much lower, and Peru is a third world country.

>And if the argument becomes "b-but we can have a counsel cherrypick phenotypes and assign weights to them based on some arbitrary perception of societal benefit", fuck outta here bud, you're becoming the pseudoscientific sjw that believes in more than 2 genders based purely on perception of the world and one's own desires clouding their judgment.

Don't talk about pseudoscience you race denying faggot. You people are a cancer, and I wish there was a purge of all race deniers from science ala what Stalin did to geneticist. Pretty interesting how communist really hate genetics don't they? But is it any surprise? Race denial is just a mask for an even worse type of denial. Gene denial. The idea that genes have 0 impact.

>>9206366

>>9214677
Can you link the articles or videos from Alternative Hypothesis that you believe define racial science?

>>9206366
Nope. Just populations.

>>9214492
fuck off kike

>>9214677
>incas
>japan
Dumb eurangutan.
Europeans got literally 1 disease called bubonic something and 40% of their population died, destroying stability on europe. Nomads invaded europe at that time and europeans waited for the mongols to collapse against reduct points that actually were less important for mongols.

Incas got 60-80% genocided, then suffered 60+% death rate epidemics due to 20+ diseases they got from pestilent eurangutans for 2 centuries. Heir princes and kings died alongside militar elites and priests due to the pestilence every time. Incas adopted iron metql woeking and started using horses less than 5 years after europeans reached South america. Even then they managed to resist for decades and secured treats for the priviledged indians. The priviledged amerindians have been assimilated by eurangutans meanwhile the low class has been living for 4 centuries mining on toxic metal quarries where the life exp. was less than 30years old. The high potential ones got into the priviledged indian class where now is the high class of Peru alongside some spanish families.

If the slavery and social selection process wasn't a problem, why do subhuman eurangutans fear that treatment?

Amerindian superiority is a historical fact. Get over it, chimpo.

>>9214798
>Also keep this thread /pol/ free.

>>9214492
Is that all you have?

>>9214798
So, I'm Jewish for citing a source.

w e w

>>9214594
It's the first thing it says:

>The proportion of human genetic variation due to differences between populations is modest, and individuals from different populations can be genetically more similar than individuals from the same population.

Or in simpler terms

>genetic variation in humans due to the different human populations isnt that great, and in some cases people from different populations can be more genetically similar than compared to those in their own population.

Otherwise saying:
>humans have so little genetic variation, people from complete opposite sides of the planet can be more similar in their genetic makeup than compared to their neighbour.

>>9215413
I don't know what more you want, it's literally the source of that bit of info.

>>9214842
Low iq Indians lost like 90% of their population in the Americas.

>>9215463
>modern comparison
So you agree with Amerindian superiority? Then if you want to compare both modern populations then eurangutans have to get 90% genocided, 60-80% epidemics for 2 centuries, high class and middle class murdered and the survivors assimilated by an inferior race, the low class thrown to toxic mines where the life expectancy is less than 30years old while being mongrelized by another inferior race and their culture is almost completely obliterated as well as the social treatment despises their population. Don't you agree?

>>9215705
literally don't know what the fuck you're talking about

>>9206366
can they mate = are they a race

your variables include

yes, no

>>9215705
Stop farting on the internet bro.

>>9215454
t1=(a,a,a,b)
t2=(c,c,c,b)

One element from tuple t1 (element b) is more similar to one element from t2 than any other element from t1. But is there more variation between or within each tuple?

>>9211495
lol you're stupid.
>The genetic variation between populations of humans is LESS than within, meaning that two European white people on average would be more genetically dissimilar than a white person and an African.

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1893020/
>Thus the answer to the question “How often is a pair of individuals from one population genetically more dissimilar than two individuals chosen from two different populations?” depends on the number of polymorphisms used to define that dissimilarity and the populations being compared. The answer, equation M44 can be read from Figure 2. Given 10 loci, three distinct populations, and the full spectrum of polymorphisms (Figure 2E), the answer is equation M45 ≅ 0.3, or nearly one-third of the time. With 100 loci, the answer is ∼20% of the time and even using 1000 loci, equation M46 ≅ 10%. However, if genetic similarity is measured over many thousands of loci, the answer becomes “never” when individuals are sampled from geographically separated populations.

>>9216170
I'm just simplifying the abstract. Honestly, I think that population isolationism will cause two individuals in two different populations to have different genetic traits, but the whole issue is bogged down with social science and anthropology. There's a fear of the possibility of taxonomic classification of race, and the division it could create. But to me, the word race doesn't really make sense to use, as we are 99.5% genetically similar to each other, and the phenotypical traits could just be adaptations to the populations environment.

>>9216190
After enough loci don't you start getting high rates of dissimilarity within populations as well?

>>9216218
chimps and bonobos share 99.6% of their DNA

>>9206619
Aaand this beautiful post goes mostly ignored for the rest of the thread while people keeps shitposting about their personal feelings when it comes to black people.

>>9206619
>Race is a pre-genetics approach to biology.
Taxonomy is a pre-genetics approach to biology

>they do not ask them about ancestry or perform any sort of biological test because such approaches have been proven garbage
>ancestry proven garbage
>DNA testing is garbage

>>9206895
What is race?

>>9210454
>Jews are caucasians
Wonder Woman is not white.

>>9206366
Pic related are all the races

>>9216190
>https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1893020/
The fact that, given enough genetic data, individuals can be correctly assigned to their populations of origin is compatible with the observation that most human genetic variation is found within populations, not between them. It is also compatible with our finding that, even when the most distinct populations are considered and hundreds of loci are used, individuals are frequently more similar to members of other populations than to members of their own population. Thus, caution should be used when using geographic or genetic ancestry to make inferences about individual phenotypes.

Lol you're stupid

Can't even read your own links

>>9216613
Lots of taxonomic groups were upended by genetics, especially in botany.

>>9209418
You've never met a German Shepard then

>>9216613
Both are pre-genetics approaches to biology, but race is worse. Those approaches have been proven garbage for making inferences about race since ultimately they don't have much to do with it. This isn't because those things are garbage but rather because race is garbage, for all of the other reasons listed in the post.

>>9214677
>Why the fuck is Lewontin's fallacy STILL being posted in 2017?
Most people aren't familiar with population genetics.

>>9206366
Humans are a self domesticated animal, they tend to form ethnic groups where people marry within the group and eventually a partially inbred gene pool appears where the members of an ethnic share certain genes. Obviously ethnic groups exchange mates and invade but in ancient times this was restricted to geographically near places. Thus the different ethnic groups of Europe, Asia and Africa are related to one another forming major races,

>>9218307
Ethnicity includes a bunch of other stuff unrelated to "race" like language, culture, and cuisine. There is research in studying how collections of ethnic groups are related and what exactly comprises ethnic identity. Ethnicity is an even more complicated concept than race but it has the benefit that it has only been studied as a social construct within the social sciences so it doesn't get all these /pol/ brainlets thinking it's biological (unlike race which used to be considered biological back in the pre-genetics era).

Also, read the thread, retard.

>>9218339
t. Angry university cuck.

>>9206366
It does, but the last time Biologists made that public it resulted in the death of tens of millions of people who got killed by other people who did not really understand that concept. So it is better to pretend they don't exist for humanity's own sake.

>>9217542
notice:
>hundreds of loci
>hundreds
this is why you're an idiot. because:
>However, if genetic similarity is measured over many thousands of loci, the answer becomes “never” when individuals are sampled from geographically separated populations.
>thousands

>>9218339
Race and ethnicity just means groups of ancestrally related people who usually share a common homeland and history.

>>9218352
Human ethnic groups, especially ones from their local habitats without a history of intrusions, are very inbred. Only race/gene denial cucks deny the level of of relatedness within a same race. Individual variation exists but racial variation is just as large.

>>9218360
>very inbred
no that's not true. there is more variation within groups than between them, but the variation between them makes them distinct on the group level.

it's like:

{AABBCC, AAbbCC, AAbbcc, AABBcc, AABbCc, AABbcc, AAbbCc}

and

{aaBBCC, aabbCC, aabbcc, aaBBcc, aaBbCc, aaBbcc, aabbCc}]

most variation is within the sets, but they can still be delineated.

>>9218370
tho it's not like it's necessarily homozygous, this was just to illustrate a point. rather you have hundreds of alleles for a gene and some pop has 20 of the variants, another has 10, etc.

>>9218370
Ah, interesting. So if you were to double-count the evidence that makes the groups distinct, you can say they're as different from the other group as the size of their entire group.

Amerindian superiority compared to europeans is a historical fact. That's for sure.

>>9217982
>>9218339

What is race?

>>9218347
t. Social media educated brainlet.

>>9218468
>>9218356

Initially race was just another term for ethnicity or nationality. Then during the 1700s/1800s the term was formalized through the advent of physical anthropology (an early pre-genetics attempt at classifying humans, scientifically). Around the same time taxonomy was being developed (as another early pre-genetics scientific field). Later in 1859 Darwin's on the Origin of Species was published. Together all of these things helped redefine race as a scientific concept (an early and ultimately incorrect one) whereby an individual's race was described primarily through physical means and believed to be derived through ancestry with races being sometimes used as synonymous with "varieties within a species".

The term remained vaguely defined and highly politicized for much of its time. In the US alone there were many supreme court cases where individuals would argue that they fit the definition (since only the "white race" could own land) only to have the Supreme court make extremely vague arguments along the lines of "no, white doesn't mean that, it means like 'you know what I mean', right?".

Once genetics took off at the end of the 1900s and the human genome was mapped in the early 2000s it turned out that a lot of early beliefs about race, evolution, and taxonomy were simply wrong. Some of the reasons are explained here: >>9206619 >>9206634 but there are many more. Moreover other results from clustering and convex analysis have shown that the problem with race isn't fixable from a mathematical perspective.

As a result two things happened:
>Population Genetics was developed as an abstract framework to study genetic distance between rigorously defined groups in general (note: traditional races are not rigorously defined and no longer a thing, though /pol/ brainlets often confuse populations for races).
>Social sciences salvaged race as a social construct (a socially defined non-scientific phenomenon).

>>9206366
Google phrenology.

Thats basically one of the key stones of race.

There are local variations like with many species but nothing that fits the distinction of race.

>>9218785
>doesn't know that phrenology is a pseudoscience

lmao

>>9218708
> convex analysis
where has this been applied?

>>9218787
>implying
http://www.bbc.com/news/technology-41188560

>>9218787
Thats the point you chuckle fuck.

If your cornerstone is made out of bullshit then the race thing is pretty bullshit.

Its a good start for a paper on how its bullshit.

>>9218790
Cluster analysis. In particular issues with the vague definition of cluster (hence many different algorithms), convergence to local maxima only (global maxima basically impossible to even check without iterating through all possible clusters, which is basically impossible), and other shit really limit the sort of assertions you can make about clusters you find in data.

The tl;dr is:
>If you have already proven that certain clusters exist in your data and you have a means of testing correctness of your algorithm then a clustering algorithm can help you find those clusters (and you can tune it to perform better).
>If you have not proven that clusters exist then a clustering algorithm can produce clusters but they may not necessarily exist. Moreover since you can't test it's correctness then different algorithms and different configurations can yield many totally different clusters (and numbers of clusters).

In other words: you can't take a dataset consisting of genetic data for a bunch of humans, run a clustering algorithm on it, and then assert that races exist because you found clusters. In fact, there are bloggers that attempt to do this (/pol/ users used to link to them in /sci/ threads) and in order to even get it to work they actually add in a bunch of custom configuration to guarantee that the algorithm will find the sorts of clusters they hope it will (of course /pol/ users cant into machine learning so they don't see that shit).

>>9218791
These guys used a Convolutional Neural Network called VGG-Face that was originally trained for facial recognition to extract features (4096 "scores" per image) from ~300k images of ~75k individuals (many individuals had more than one image). VGG-Face's scores are difficult to interpret but it looks at everything and likely contains information for predicting stuff like grooming (including facial hair), tanning, makeup use, etc... They trained a logistic regression algorithm over 19 sets of data (20th for testing).

Ultimately they asserted that VGG-Face can be used to identify gay people and tried to offer some explanations about why (eg. grooming). Truth is we should really be far more concerned about the effectiveness of facial recognition DNNs than we should about these guys using it this way. Greentext for emphasis:
>If they could do it with a shitty technique like logistic regression over such a small data set obtained through a dating site using only VGG-Face for feature extraction then just about anyone can do it and it can be done a LOT better!!!

Note: They did not use the technique to assert the existence of gays, which would be more analogous to the whole "race exist becuzzz ----" argument.

>>9218708
> rigorously defined groups

Is "species" rigorously defined? There are plenty or examples of cross-species, fertile hybrids.

>>9218785
>phrenology is one of the key stones of race

Huh? No it isn't.

>>9218876
>They did not use the technique to assert the existence of gays, which would be more analogous to the whole "race exist becuzzz ----" argument.

So gays don't exist?

>>9218866
so this is about you need to see if things fall into your defined categories to use f-score, rand index, etc. and since hierarchical clustering has a whole host of different linkages and methods that will yield different results you can't say the clusters found have any meaning?

you can see how well different algos operate on toy examples then just apply them irl. they do this a lot in finding operational taxonomic units for a whole host of other species, idk why that is such a problem here.

>>9218866
so this is about you need to see if things fall into your defined categories to use f-score, rand index, etc. and since hierarchical clustering has a whole host of different linkages and methods that will yield different results and we currently have no means of optimizing them you can't say the clusters found have any meaning?

you can see how well different algos operate on toy examples then just apply them irl. this process should be p close to rigorous optimization. they do this a lot in finding operational taxonomic units for a whole host of other species, idk why that is such a problem here.

>>9218866
so this is about you need to see if things fall into your defined categories to use f-score, rand index, etc. and since hierarchical clustering has a whole host of different linkages and methods that will yield different results and we currently have no means of optimizing them you can't say the clusters found have any meaning?

you can see how well different algos operate on toy examples then just apply them irl. this process should be p close to rigorous optimization. they do this a lot in finding operational taxonomic units for a whole host of other species, idk why that is such a problem here. hell there is SSE and wilk's A, etc that can be applied too.

>>9218961
>So gays don't exist?
I only said this method isn't sufficient to assert their existence (though lots of other unrelated research says they do) because it assumes they exist as a precondition, such a claim would be circular logic.

>>9219008
How can you even have "defined categories" to test without assuming your categories exist in the first place? Moreover if you assume they exist then you can't use it to assert existence because that would be circular logic. I was talking about clustering algorithms in general, not merely hierarchical clustering, but yes that's part of it. Two other issues are that the clusters found don't resemble traditional races unless one tweaks the data/algorithm (circular logic) and that in general on what basis can we even assert the clusters have meaning if we don't already know they do?

>>9219033
>How can you even have "defined categories" to test without assuming your categories exist in the first place?
you would make toy examples of several kinds of distributed data points and see which algo recovers them, do bench marking and go with the best algorithm.

>Moreover if you assume they exist then you can't use it to assert existence because that would be circular logic.
isn't a big data technique to use clustering to see if there are any clusters in the data? like use k-means then check when SEE is minimized to find clusters in the data?

>I was talking about clustering algorithms in general, not merely hierarchical clustering, but yes that's part of it.
in taxonomy since they're using distance matrices of sequence pairs it's typical hierarchical or some kind of heuristic or mix of them.

>the clusters found don't resemble traditional races unless one tweaks the data/algorithm (circular logic)
why tweak the algorithm? why not just use those clusters to redefine the categories?

>and that in general on what basis can we even assert the clusters have meaning if we don't already know they do?
why can't the most reliable algorithm applied to the dataset with the minimum SEE or whatever, those clusters be the meaning? why can't race be defined as full genome intraspecies clusters?

>>9219057
>you would make toy examples of several kinds of distributed data points and see which algo recovers them, do bench marking and go with the best algorithm.
Unless I'm misunderstanding you here I think you're missing the point. Would you care to elaborate?

>isn't a big data technique to use clustering to see if there are any clusters in the data? like use k-means then check when SEE is minimized to find clusters in the data?
Not exactly. Clustering is classified as an unsupervised learning technique under machine learning which means that there are no labels and no testing set at the end. These algorithms will ALWAYS "find clusters" (it will never fail and say clusters don't exist), it just won't always be clear that your clusters are meaningful in the way you want them to be (or in any way). Moreover, running the same algorithm on the same data can yield totally different clusters (local maxima) and there's lots of configurations and algorithms one can use. So typically one will do a lot of iterations and tweaking to find desirable clusters (SSE is part of that process, actually finding the global maximum is not only impossible but may not be desirable).

>in taxonomy....
I agree that hierarchical makes more sense here but I was speaking in generality. Besides the /pol/ blogposts I'd seen in the past used k-means or something like it.

>why tweak....
Good questions. I think that is what one ought to do (assuming they've actually found a global maxima). Unfortunately I suspect that /pol/ and other brainlets would get buttblasted over people redefining something central to their ideology.

>why can't the most...
"The most reliable algorithm" isn't really a thing. That's kind of the point. Also, suppose an ayylmao gave you a blackbox for finding the optimal clustering on any data set and you fed it white noise as data. Would the clusters obtained have any meaning?

>>9219133
>Clustering is classified as an unsupervised learning technique under machine learning which means that there are no labels and no testing set at the end.
right you can't use f-score, rand ind, etc because you need to know the labels to get TF, TP, FN, etc. i get that and

>These algorithms will ALWAYS "find clusters"
and that too, you set k =3, there will be 3 clusters even if 2 clusters have single datapoints. but isn't that were SSE and other validity measures comes in?

given your example of white noise, whatever pattern it finds wouldn't that cut off at 0 if it is truly distributed randomly and not that not really random rand()?

also:
>actually finding the global maximum is not only impossible
you absolutely sure about this tho? is there a proof?


also given there is no good method for intraspecies clustering should we just use pca then?

>>9219245
>but isn't that were SSE and other validity measures comes in?
Yes, those are some of the goals of validation (and SSE is one of its tools), but validation is super hard and will sometimes force you to make subjective decisions or debate dumb philosophical questions.

“The validation of clustering structures is the most difficult and frustrating part of cluster analysis."
-Algorithms for Clustering Data, Jain and Dubes

Here are some slides:
Skip to slide 33 for general concerns with clustering and 42 for an overview on clustering validation techniques.
https://www.cse.buffalo.edu/~jing/cse601/fa12/materials/clustering_basics.pdf
Note: External validation is what I was referring to as circular logic with regards to race. So that approach cannot be used to assert the existence of race.

>wouldn't that cut off at 0
I'm not sure what you mean by this. See pic related though for some examples of clusters in random data. One of the goals of clustering validity is to avoid finding clusters in random data (the slides above have more info). It's not an easy thing to do though as I describe above.

>you absolutely sure about this tho? is there a proof?
My understanding is that currently the only way to achieve it is to actually test all possible clusters and that sort of thing isn't technically impossible but it is totally intractable (the reason we use these algorithms in the first place is to avoid that). I suppose it's possible that sometime in the future someone could make a really clever breakthrough making it feasible to do. Also, in some cases where external validation is available you might be able to pull it off too.

>PCA
Unless I'm missing something, PCA, will just try to reduce the dimension of your data set (though you'd still likely end up with hundreds or thousands of dimensions).

>>9219366
those are some good slides, anon. thanks.

>External validation is what I was referring to as circular logic with regards to race.
yeah we are lacking the ground truth here if there is any. initially i was on about mock clusters and doing some external validation to get some benchmarks but nvm that. it's more that when there genuinely is internal validity it can't just be dismissed. no matter how much it negates our preconceptions of what we think the clusters are. we're looking at sequence differences, not even what amino acid may result from them. and if there is internal validity what more is there?

>One of the goals of clustering validity is to avoid finding clusters in random data
that's exactly what i am getting at, if we do internal validation correctly and i'm not trying to say it's easy shouldn't we know when there are no clusters there even as it will show 3 clusters at k=3? those validity measures should be able to pick up on white noise.

>Unless I'm missing something, PCA, will just try to reduce the dimension of your data set (though you'd still likely end up with hundreds or thousands of dimensions).
yeah it's just for visualization. exactly what it is, it's eigenvalue decomposition of a contrivance matrix. i'm just thinking instead of relying on clustering if we are having the problems telling groups of people apart without any ground truth, pca is a useful tool because you can see if people map on geographically even if there is no reliable clustering.

https://faculty.washington.edu/tathornt/SISG2015/lectures/assoc2015session05.pdf

>>9206366
It does, but it is a useless classification. Too much crossbreeding going on for it to be useful.

>>9206619
>>9206634

welp, this thread should have stopped here. sorry racists

here is a 3D PCA plot of human genetic diversity

the fact that there are clusters and these clusters correlate with geographic distributions and roughly even with classical ideas about race conclusively proves that a certain biological notion of race is definitely real

it is not race as imagined in the past, meaning neatly distinct groups (in which case we would see several distinct blobs)

but it is also not true that humanity is all the same and no major genetic clusterings exist (in this case we would see one single blob or alternatively, random spread of colors along the plot)

so the truth is somewhere in between what the racists say and what anti-racists say

>>9219417
>contrivance
lol that should've been covariance. fug stayed up way too late finishing up this code and shitposting.

>>9219417
>that's exactly what i am getting at, if we do internal validation correctly and i'm not trying to say it's easy shouldn't we know when there are no clusters there even as it will show 3 clusters at k=3?
Note quite, you would have to make an ultimately subjective choice. More than likely it will find some crappy clusters where it's not super clear if they are cluster or not and it's not super clear if they're meaningful or not (eg. maybe the clusters will become more clear with more data but right now it's just too sparse).

>those validity measures should be able to pick up on white noise.
White noise is an extreme example but even then all it really tells you is that those clusters suck.

>PCA and those slides
Okay, I understand now what you're referring to with PCA. I didn't know people were applying it to ancestry and population in that way. Seems totally wrong to me since by focusing on just a couple PCs they're ignoring tons of dimensions in their data just because it's convenient for graphing on a 2D plane. Who knows though, I haven't looked at their paper, maybe they've got some argument involving significance for the things they want to select that justifies them choosing those PCs or something. I also have some concerns over how they define ancestry and justify using PCs to make inferences about it, but it's possible that's resolved in some way unrelated to this technique.

On a more personal note I don't think PCs are an appropriate technique for this sort of thing since you end up ignoring so much data for the sake of a simple visual. Autoencoders seem much better, though I don't think stats people have really caught onto them.
https://danluu.com/linear-hammer/

>>9219472
Part of the argument is that human genetics contains tons of information and only a tiny part of that manifests as the physical characteristics that we associate with race. Race in the best case scenario misleads us into ignoring the majority of human genetic information. Moreover, there is really no reason why one can't create a classifier that ignores most of one's genetic data and only pays attention to the small portion associated with race. Arguably that's what one is doing when they use PCA and choose PCs that identify the data they're interested in.

At the same time, focusing only on a few PCs (and arguably using PCA in general) can cause you to accidentally merge distinct clusters or to artificially separate the same cluster. See pic related for a crappy example I drew really quick. In blue are your PCs. Now, throw out the long one and scale the small one and you've got two distinct clusters instead of one big one (kind of, my picture is not very representative of real data). Now consider that in your picture they're ignoring hundreds or thousands of PCs (projecting down to just 3) and one really has to begin to wonder what the data really looks like.

Here is another example showing how PCA can merge clusters.
https://www.quora.com/Does-it-make-sense-to-perform-principal-components-analysis-before-clustering-if-the-original-data-has-too-many-dimensions-Is-it-theoretically-unsound-to-try-to-cluster-data-with-no-correlation/answer/Maruti-Agarwal

>but it is also not true that humanity is all the same
Literally no one on /sci/ ever says that. It is a /pol/ strawman. The argument is that there isn't a "standard" partitioning of humans because the data is too messy to find or justify one. Population Genetics is a framework for studying genetic distance between groups in general and the way things are done now.
>and no major genetic clusterings exist
It's sufficient to point out that all the clustering offered are garbage.

>>9219496
it's where most of the variance lies in the dataset. that's p useful. and sure most of the variation is between groups but the clearly the biggest difference is between them.

>>9219545
I think you misunderstood. Though PCA can be bad in some cases that wasn't the point I was making. Rather I was saying that applying PCA and then only looking at a few PCs (while ignoring the rest) is bad unless you've got a really good reason to do that.

Check this out as an example:
http://gallantlab.org/huth2016/
They used PCA to do something similar to what you're arguing. Initially they began with a large dataset and through PCA they reduced it down to a 985 dimensional space. That's 985 Principal Components. Now, because the data they were dealing with was inherently noisy and came from different they were able to deduce that all but 3 of those PCs could be explained by noise for all of their subject's data (though a 4th PC seemed significant for all but one subject). Those 3 PCs were mapped to an RGB color space and they were represented in the data directly. The fourth PC is visualized in their paper and set as the radius for the color wheel on the bottom left of their visualization.

In this case they ruled out dimensions methodically by relying on [external] facts they knew about their data.

On the other hand I see no similar justification used for ignoring PCs in the ancestry/genotype stuff. Instead it seems like they had already decided ahead of time what they were looking for and merely found a way to get it. This is fine but can't be used to assert existence due to circular logic.

>>9219568
well if you're trying to see where the biggest variance lies, that's all that needs to be done. look, senpai. the fact that the biggest variance fits with ancestry is pretty telling. we're looking at sequences and trying to see if there are patterns in the data, there are, and the most noticeable patterns coincide with ancestry. that's not to say that is all the pattern there is, just pc1 and pc2. and unlike with mri images and any signal processing stuff there is no noise to worry about. that's in the sequencing stage and the high throughput stuff deals with that fine. it's not making associations with gene expression or anything where noise is a problem.

>>9215864
>>9215770
I just made an observation about Amerindian development rate being superior to europeans, then some eurangutan mentioned some assumption measures from modern amerindians, so I proceeded to remind him how to compare two entities (same circumstances). So I listed the ammount of social selection and changes amerindians suffered and that eurangutans had to receive the same treatment to be compared to superior Amerindians.

How is this hard to get?

>>9219600
How are you not understanding?

Brain scenario:
>Data set with tens of thousands of dimensions.
>Apply PCA to reduce the number of dimensions to 985.
>Measure PCs to see which ones are significant and which ones can be explained by noise, only 3 significant PCs remain.
>Use 3 remaining PCs to draw inferences from.

Genotype scenario:
>Data set with 200,000 dimensions chosen by unknown means (~150m SNPs in humans).
>Apply PCA to reduce to the number of dimensions to some still very fucking large number.
>"unlike with mri images and any signal processing stuff there is no noise" so all dimensions are significant
>Arbitrarily decide to only look at first two dimensinos/PCs and run to the publishers once they satisfy my biases.
>Ignore all other PCs.

Yes, PCs store the "biggest variance" because that's how PCA works.
Yes, the PCs are ordered from biggest variance to smallest.
No, the two first PCs don't tell you everything you need to know about your data.
No, for making inferences about the existence of clusters it isn't sufficient to take a high dimensional data set and project everything down to two dimensions (because in case it wasn't clear: when you ignore a PC you are performing a projection of that axis) and call it a day (cf. internal verification + merging clusters + creating clusters).

To use an analogy:
>You are sitting in a cave.
>Outside the cave a 150+ million dimensional creature passes by.
>Some small amount of light hits the creature in such a way that 200,000 of its dimensions are distorted and reduced in number before being projected as a 2 dimensional shadow on your cave wall.
>You jump up in chimp like excitement as you howl madly into the night in joy because at last you "truly understand" the creature's true nature.

>>9219643

>No, the two first PCs don't tell you everything you need to know about your data.

They tell you a lot because those are the most significant dimensions. And when they coincide with ancestry/ethnicity data then it tells you that those are significant, too.

How significant is most significant is another matter.

>>9219643

>Arbitrarily decide to only look at first two dimensinos/PCs

>first two dimensions

>arbitrary

I dont think you really know how PCA works..

>Ignore all other PCs.

You can ignore them to get a high level overview of the dataset. Look at that 3D PCA plot

>>9219472

now mentally squash PC3 to get a 2D PCA plot. You lose some details but high level features of the dataset are still clearly visible. Same analogy applies to squashing a 200,000 dimensional dataset to 3 dimensions using PCA. It does tell you quite a bit about the dataset.

>>9219662
>>9219651
The issue I'm pointing out is far more pedantic than that.

I could take a bunch of genetic data and use a similar approach to try and find clusters of people with different hair color. It might even be sufficient to use a single PC and with even less SNPs as data with the same process as the genotype technique. Could I argue that my technique allows me to identify people with different hair color? Sure, it clearly does. Could I argue that it is okay to ignore other PCs in my approach? Sure, my job isn't to prove the existence of hair color, only to develop a means of predicting it from this data set.

On the other hand, could I take my very simplified model and say:
>Because my model is significant, I can prove that THE human races (i.e. THE clusters in human genetic data) are exactly "blonde", "brunette", and "redhead" as predicted in my model.
No, I'm ignoring so much data that it's obvious that THE human clusters are likely to be far more complex. Sure, hair color might be a factor but it's likely there's a whole lot more to it especially as the number of SNPs goes up and less PCs are ignored (note: it isn't uncommon to apply PCA to a dataset BEFORE running all PCs through a clustering algorithm).

That's exactly my argument here. This cannot be used to make grandiose claims about THE human clusters (i.e. races). It can only be used to deduce ancestry from the data provided you already know what ancestry is (rigorously).

>now mentally squash PC3 to get a 2D PCA plot. You lose some details but high level features of the dataset are still clearly visible. Same analogy applies to squashing a 200,000 dimensional dataset to 3 dimensions using PCA. It does tell you quite a bit about the dataset.
Refer to problems pointed out in >>9219518 . Note that these problems don't say that your dimensions don't hold significant data, but they do say that the clusters you're observing could be totally different when all other PCs are taken into account.

>>9218339
>includes a bunch of other stuff unrelated to
Said no scientist ever.

>>9219643
>How are you not understanding?
how are you not understanding my point?

>Measure PCs to see which ones are significant and which ones can be explained by noise, only 3 significant PCs remain.
this needs to be done because of noise. not an issue when just looking at clean sequences.

>Data set with 200,000 dimensions chosen by unknown means (~150m SNPs in humans).
just randomly selected snp micrtoarrays, so no bias.

>Arbitrarily decide to only look at first two dimensinos/PCs and run to the publishers once they satisfy my biases.
nope, select the first two because that's where the biggest variance is on to see what dimensions are most significant. again no noise.

>No, the two first PCs don't tell you everything you need to know about your data.
which is what i said.

>No, for making inferences about the existence of clusters it isn't sufficient to take a high dimensional data set and project everything down to two dimensions (because in case it wasn't clear: when you ignore a PC you are performing a projection of that axis) and call it a day (cf. internal verification + merging clusters + creating clusters).
wasn't what i was getting at. i was merely saying pca for looking at distribution of the dataset. seeing how the data is distributed. and it is distributed very strongly ancestrally.

>>9220307
in fact ancestry is demonstrated to be the main distribution pattern, pc1.

>>9220307
>>9219643

what are you guys even arguing about? if you already have clearly defined groups over the first two principal components, then the raw data will also have clusters which are at least as distinguishable from one another. pca will never increase the distance between two points.

>>9220316
if i am getting his argument right he is saying picking pc1 and pc2 is arbitrary. and he's using a fmri study which had to deal with noise but there is no noise to deal with here. no need to associate p-values with pcs. the most significant pcs are the first ones.

>>9220330
>the most significant pcs are the first ones.

hence the name "principal" component analysis.

>>9220343
yep, tho i see his point when there is noise it's not so simple but that's not an issue here.

>>9220346

noise does not add extra dimensions to your data, and if the variation along your first principal component is mostly due to noise then PCA is probably not the right tool for the job.

>>9206366
There is no race DNA based

>>9209418
Your dog analogy was retarded and wrong.

>>9220348
yeah, noise won't add dimensions. the dimensions are of the cov matrix, but his example was finding the noisey pcs and removing them.

>>9220363

if you already know that the variation along a certain direction is due to noise, can't you remove that without doing the usual PCA algorithm?

pca is generally not a tool for reducing noise afaik

>>9206504
In school we measured bone lengths and structure and we found that the bone lengths differed in the humerus bones of asian people, white people, and black people. Visually shapes of the skulls are anatomically different. There is a physical difference and you can tell which skull is which once you learn the characteristics. Though there is a slight percentage you can get a race wrong by just using one technique.

>>9220307
>wasn't what i was getting at. i was merely saying pca for looking at distribution of the dataset. seeing how the data is distributed. and it is distributed very strongly ancestrally.
Sure, okay. So you found a correlation, for something that you could've obviously found a correlation for.

>>9220330
That's not it at all. The issue is that the shitty way you're going about it is at best meaningless and at worst circular logic.

>>9220316
Performing PCA and then throwing out a lot of dimensions can make data appear to be clustered when it isn't actually clustered (i.e. it would perform badly with regards to SSE and other cluster validation tests). That said, anon doesn't care because they apparently aren't interested in trying to find the actual clusters that humans fall into with total logical rigor. Rather they're just interested in finding a correlation and applying circular logic to claim it means something.

>>9220346
What the fuck are you talking about? You obviously don't get my point.

>>9220348
>>9220363
I never said noise adds information. I have no idea where that anon is getting that.

>>9220422
That's not how PCA was used in their paper.
>They collected data from a bunch of patients over a few different sessions.
>They used PCA to reduce the dimensionality of the data because it was fuckhuge.
>This left them with ~1000 dimensions PCs.
>They then tested the PCs against data from each subject to see which PCs had statistically significant variation (i.e. couldn't be chalked up to noise).
>Unfortunately this left them with only 3 usable PCs (4 if you ignore one subject).
>They then used all 4 remaining PCs to try and draw inferences as best they could.

The only reason I brought up that study was to point out how someone who is actually approaching the data without any preconceptions would try to salvage as much as possible in order to find as they can.

>>9220448
>Performing PCA and then throwing out a lot of dimensions can make data appear to be clustered when it isn't actually clustered (i.e. it would perform badly with regards to SSE and other cluster validation tests).

true, but SSE is not necessarily a good cluster validation test. and actually, i think pca would favor SSE since it minimizes the reduction in variance.

>>9214219

I think when it comes to genetics it depends on how you look at it.

I mean, consider two sets, one consisting of all integers and one consisting of all reals except for the integers. You would always find a real that's closes to some arbitrary integer than any integer. But that doesn't mean reals are closer to some integer n than n+1. After all, you can always find some non-integer real that's closer to an integer than the one you have. The same doesn't apply to integers.

In humans, it doesn't.

t. actual biologist

>>9219518
>only a tiny part of that manifests as the physical characteristics that we associate with race
only a tiny part of that manifests as the physical characteristics that we associate with species

>>9220474
>Performing PCA and applying a clustering algorithm on the resulting data set: Yes, this is fairly common.
>Performing PCA, defining clusters by just looking at the first two principal components, and then using cluster validation techniques (including but not limited to SSE) on the data across all PCA dimensions: No, ignoring PCs projects data onto a lower dimensional space and brings points closer together than they actually are (which is part of what SSE tests against, namely "are the samples in your cluster actually close to each other").

There isn't a perfect one size fits all solution for cluster validation. Instead you've just got a bunch of tools and you have to be smart about using them. Some good intro slides were linked in >>9219366

>>9220448
>That's not it at all. The issue is that the shitty way you're going about it is at best meaningless and at worst circular logic.
right picking the dimension with the largest variance when there is no noise to deal with is shitty how exactly? i'm not saying there are no other dimensions as you seemed to have implied i'm saying, just this is the one where most of the variance lies.

>Performing PCA and then throwing out a lot of dimensions can make data appear to be clustered when it isn't actually clustered
but i'm not talking about clustering anymore, just looking for how variation is distributed and it so happens for pc1 it coincides with ancestry. this shouldn't be a problem unless you don't like the finding.

>it would perform badly with regards to SSE and other cluster validation tests).
unless that is done you cannot just assume it will score badly as you have repeatedly claimed. they can't say their clusters are meaningful either unless they validate them. like it doesn't look like noise. it maps on to geography if anything. i haven't seen a study looking at SSE, etc for these ancestry clusters. if you have pls share. the only validation i have seen was external:
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1196372/
>Of 3,636 subjects of varying race/ethnicity, only 5 (0.14%) showed genetic cluster membership different from their self-identified race/ethnicity.
seems p good you ask me.

>They then tested the PCs against data from each subject to see which PCs had statistically significant variation (i.e. couldn't be chalked up to noise).
yes they looked for the pcs with the lowest p values because they were dealing with fmri signals which are noisey so they had to do this to remove the pcs which were noisey. leaving them which pcs which aren't.

but noise isn't an issue here so the most significant pc is the first one. since again we are looking at sequences and how they are divergent from one another.

>>9220492
Yes, this is true. It's also why we've seen the field constantly make an effort to reclassify things based on genetics (obviously still a work in progress).

>>9220494

what i said in >>9220316 pretty much still applies. say that you're looking at the first few principle components, and you see two distinct groups of points. these will likely form two distinct groups in the original space. if there was enough variance along the discarded component that a single group now appears as two separate groups, then that direction would probably have been the first principle component. right?

>>9220499
>I just chose a dimension that's at least as big as all the others. How is that not the best decision?!
lmao

A correlation doesn't say any meaningful about the existence of race which is the point of this thread. Such studies aren't performed because those papers aren't interested in finding better more interesting clusters, they only care about predicting ancestry.

>self-identified race/ethnicity
When you've found yourself desperately turning to social science for your external validation set nothing is "p good".

>yes they looked for the pcs with the lowest p values because they were dealing with fmri signals which are noisey so they had to do this to remove the pcs which were noisey. leaving them which pcs which aren't.
So what is your excuse for ignoring data? It seems the obvious answer is that you don't actually care about everything the data has to say and are only interested in hearing what you want.

>>9220524
You've got it backwards son.
>say that you're looking at the first few principle components, and you see one big blob. Is it not possible that from the perspective of some of the PCs you ignored the blob is actually two or more clusters?

For a simple example refer to
https://www.quora.com/Does-it-make-sense-to-perform-principal-components-analysis-before-clustering-if-the-original-data-has-too-many-dimensions-Is-it-theoretically-unsound-to-try-to-cluster-data-with-no-correlation/answer/Maruti-Agarwal

>>9206366
You could say that the genetic variation between individuals of a single race is greater than the variation between individuals of different races.

>>9219518
>The argument is that there isn't a "standard" partitioning of humans because the data is too messy to find or justify one.
What does "standard" mean?

>>9220539
>>say that you're looking at the first few principle components, and you see one big blob. Is it not possible that from the perspective of some of the PCs you ignored the blob is actually two or more clusters?

oh yes, of course. that's not what you were arguing though.

>>9220448
>Performing PCA and then throwing out a lot of dimensions can make data appear to be clustered when it isn't actually clustered

>>9220539
>A correlation doesn't say any meaningful about the existence of race which is the point of this thread.
what would for you?

>Such studies aren't performed because those papers aren't interested in finding better more interesting clusters, they only care about predicting ancestry.
and if there was nothing with the sequences they couldn't do it. seems p meaningful to me.

>When you've found yourself desperately turning to social science for your external validation set nothing is "p good".
funny how it coincides with genetics tho. the categories have an outward appearance and that has a genetic basis i don't see the problem here.

>So what is your excuse for ignoring data?
i'm not ignoring the data, i'm simple looking for where most of the variation lies and it so happens to coincide with ancestry.

>It seems the obvious answer is that you don't actually care about everything the data has to say and are only interested in hearing what you want.
fucking rich dude. you're the one who is buttblasted over the main pcs having so much to do with ancestry and are desperately scurrying to deny their validity like pol denying their african admixture.

>>9206914
Animal species with an appropriate amount of environmental diversity do. For instance, junglefowls and chickens are the same species, but they have a wide array of differences because of the different environmental pressures.

There's only one race, the human race. And everyone is invited. Except the darkies.

>oh yes, of course. that's not what you were arguing though.
Those two greentexts were supposed to mean the same thing but the latter one is kind of unclear. I should've maybe phrased it as it can make separate clusters appear as one cluster after throwing out dimensions or something. Either way, I'm glad the confusion is resolved.

>>9220571
>what would for you?
Something logically sound for starters.

>coincides
That's a strange way of saying
>A implies B but B doesn't imply A

Yes, funny how genetics is a way better set of data for drawing assertions and even with poorly defined notions like that it can yield some success. Forget that countless anons pointed it out early on ITT and it's really kind of obvious since race is just a shitty pre-genetics attempt at classifying phenotypes.

>fucking rich dude. you're the one who is buttblasted over the main pcs having so much to do with ancestry and are desperately scurrying to deny their validity like pol denying their african admixture.
No, I don't care about that result. It's simply not a very interesting result as anyone who has done any bit of machine learning would've seen it a mile away. I'm more annoyed that it could be done much better but people would rather dwell in mediocrity and the way it's done makes it stupidly easy for people to misinterpret and fall into circular logic (which is exactly the sort of shit that ends up creating the misconceptions in these threads in the first place)..

We know different groups look different and all about phenotypes, but speaking strictly mental differences... Even between genders, have been exaggerated to absurdity. Now, we see how more and more how people can be convinced they have the mind of the other gender, to explain this as anything but a symptom of culture and the obedience conditioning is a mistake, though I'm sure chemical imbalance, or being poisoned helps. As for the twin studies and other attempts to experimentally prove cognitive differences, I've read them and am not convinced by their methods. When you are looking for differences, you will see them, this doesn't mean they're there, or the cause.

tl;dc: Adaptation, you are your belief

>>9214219
He meant that the difference between the average African and the average European is smaller than the difference between two randomly picked Europeans.

>>9220637
>Something logically sound for starters.
how would you go about finding people's ancestry? since 23andme and everyone else is doing it wrong apparently.

>A implies B but B doesn't imply A
don't have to be equivalent.

>Yes, funny how genetics is a way better set of data for drawing assertions and even with poorly defined notions like that it can yield some success.
it yields pretty good success.
>0.14% error

>Forget that countless anons pointed it out early on ITT
and I was the first to even mention validation.

> it's really kind of obvious since race is just a shitty pre-genetics attempt at classifying phenotypes.
i've seen the term in population genetics, there is a textbook definition. forget what page it is on:
http://libgen.io/book/index.php?md5=FD0D8A4BA470F87AE390911E7135D624
should be on population substructure..

>No, I don't care about that result. It's simply not a very interesting result as anyone who has done any bit of machine learning would've seen it a mile away.
well then we have no problem.

> I'm more annoyed that it could be done much better but people would rather dwell in mediocrity
i agree, something that bothered me with these papers is the complete lack of internal validation. perhaps i'll get their fasta files and do it myself one day.

>and the way it's done makes it stupidly easy for people to misinterpret and fall into circular logic (which is exactly the sort of shit that ends up creating the misconceptions in these threads in the first place)..
because clustering algorithms get back clusters therefore there are races is stupid and wrong i agree. however if you see clusters form after a pca, pca may merge but they can't split iirc. also a lot of the misconceptions i was seeing was regarding how dissimilar people are genetically, the whole any african is as similar to any european as they are with another european. so really both extremes are wrong in this debate as another anon pointed out.

>>9220637
>Something logically sound for starters.
how would you go about finding people's ancestry? since 23andme and everyone else is doing it wrong apparently.

>A implies B but B doesn't imply A
don't have to be equivalent.

>Yes, funny how genetics is a way better set of data for drawing assertions and even with poorly defined notions like that it can yield some success.
it yields pretty good success.
>0.14% error

>Forget that countless anons pointed it out early on ITT
and I was the first to even mention validation.

> it's really kind of obvious since race is just a shitty pre-genetics attempt at classifying phenotypes.
i've seen the term in population genetics, there is a textbook definition. forget what page it is on:
http://libgen.io/book/index.php?md5=FD0D8A4BA470F87AE390911E7135D624
should be on population substructure..

>No, I don't care about that result. It's simply not a very interesting result as anyone who has done any bit of machine learning would've seen it a mile away.
well then we have no problem.

> I'm more annoyed that it could be done much better but people would rather dwell in mediocrity
i agree, something that bothered me with these papers is the complete lack of internal validation. perhaps i'll get their fasta files and do it myself one day.

>and the way it's done makes it stupidly easy for people to misinterpret and fall into circular logic (which is exactly the sort of shit that ends up creating the misconceptions in these threads in the first place)..
because clustering algorithms get back clusters therefore there are races is stupid and wrong i agree. however if you see clusters form after a pca, pca may merge but they can't split iirc. also a lot of the misconceptions i was seeing was regarding how dissimilar people are genetically, the whole any african is as similar to any european as they are with another african. so really both extremes are wrong in this debate as another anon pointed out.

This thread is so absurd. Arguing against race existing by screwing around with lots of big data Math is like trying to disprove gravity. Is it because we are trying to define race as some huge umbrella that determines a lot of characteristics?

How about this: skin color exists. So does eye shape. Those 2 things are what the general public uses to define 'race.' It exists. We see it every day.

You lose.

>>9220707
That doesn't really sound the same as saying "two European white people on average would be more genetically dissimilar than a white person and an African".

At any rate, according to OOA, there's less genetic variation within Europeans than Africans. So two randomly chosen Europeans should be less different from each other than two randomly chosen Africans are to each other and therefore a randomly chosen European is more similar genetically to another random European than a random African.

>>9220879
So in the interest of the species interracial marriage should be illegal

>>9209697
Do you? Hispanic is not a race, it's an ethnicity bound by coming from a spanish-speaking culture, which is a white European language originally. There are white Hispanics, black Hispanics, Asian Hispanics, any race can be Hispanic. It's just most people think mestizo when they hear "Hispanic" (European and Amerindian mix)

>>9206634
Thank you

>>9211397
I think it's likely a subset of the population that those geological areas are derived from were engineered by something else. Hominids that came before us, other species, aliens, whatever. It just doesn't seem that we arose naturally.

Maybe those holes were ports for some sort of device. Perhaps they had a direct purpose relative to a certain spatial and temporal location. Garden of Eden of course comes to mind, which is derived from the much earlier tales of gilgamesh, which was probably derived from something earlier.

Here's an explanation for retards trying to argue about the Witherspoo et al 2007 paper.

Two populations of humans can have almost the exact same frequencies of alleles of their genes. However, you can take one individual and statistically match their own alleles to the small differences in the frequencies, even if the those differences in frequencies between populations is very small. Over a large number of SNPs, you can "match" a specific (not random) individual to a group.

As the paper states here >>9217542 just because you can correlate a person to a population doesn't imply those populations are different in any way.

>>9221132
>just because you can correlate a person to a population doesn't imply those populations are different in any way.
It literally does.

>>9221244
>look at only the things that support your conclusion
>ignore everything else

>>9221244
I'll make it as simple as I can. Let's makes an example with one gene that has two alleles, A and B. Let's say population 1 has frequencies of A 50.1%, B 49.9%, while population 2 has frequencies of A 49.9%, B 50.1%.

If an individual has a allele A there is a very small statistical chance they actual belong to population 1, even those both allelle frequencies are basically 50/50 for each population. The Witherspoon paper just adds those probabilities up over a large number of SNPs.

>>9215864
Nice

>>9221132
the closer you get to whole genomes the smaller the dissimilarity fraction.
>Dissimilarity fraction ωˆ: Let ω be the probability that a pair of individuals randomly chosen from different populations is genetically more similar than an independent pair chosen from any single population. We compute all possible pairwise genetic distances, classify them as within- or between-population distances (the sets dW or dB, respectively), and then calculate the frequency with which dW > dB (that is, a within-population pair is more dissimilar than a between-population pair). This fraction, equation M1 is an estimator of ω. The expected value of equation M2 ranges from 0 to 0.5 (regardless of the number of populations). At equation M3 = 0, individuals are always more similar to members of their own population than to members of other populations; at equation M4 = 0.5, individuals are as likely to be more similar to members of other populations as to members of their own. The distributions of pairwise genetic distances implied here resemble the common ancestry profiles proposed by Mountain and Ramakrishnan (2005), who use a different measure of genetic distance. The shared-alleles distance used here generally yields slightly lower values of equation M5

the person you linked didn't read the paper.
see: >>9218352

>>9221325
>Let's makes an example with one gene that has two alleles
wow what a oversimplified example devoid of reality.

>>9206366
>We have to do a assignment in Physical Anthropology explaining why race doesn't exist biologically
In fucking Anthropology? What the fuck? Anthropologists can tell the race of skeleton easily. This has to be elaborate bait. Race is a touchy subject in the West, but ask any smarty Asian scientist and it is rather settled science.

>>9208535
>all human groups would have the same IQ with the same opportunities
They don't though. The Minnesota Transracial adoption study being the most famous.Not to mention we can compare the SAT results of rich blacks to poor whites.

>>9211495
This is moronic. It does not take into account the significance of the variation. As an example if the variation within a racial group has an effect on hair and eye colors (like how Europeans can have blond, red, brown, or black hair, and blue, brown, or green eyes) then it dosen't mean much in terms of differences. If the difference between race A and race B is neurological development, then that results in huge empirical differences.

Genetic Variance != Different Race

The TYPE of genetic variance is the important part especially if it has a notable effect on behavior, as groups that get along will have complimentary or similar behaviors.

>>9221496
>if that statistic is true then it's likely it was written by a chinese anthropologist.

>>9211495
Straight up fallacy, pic related.

>>9221528
There isn't more genetic variance within races. But the point about what type of variance is also important.

>>9221559
If you add enough loci you ought to get superior clusters that no longer resemble shitty racial classifications.

>>9221566
[citation needed]

>>9221566
wrong. the more loci the more clear the separation, see: >>9218352

>>9221582
Yes, more separation. So much so that new clusters become visible.

>>9220754
How is merging not a problem to you?

Suppose:
>You are an incompetent ayylmao tasked with studying Earth's languages. At the end of your research you conclude that there's basically two clusters of languages based on how they sound:
>1: https://www.youtube.com/watch?v=Vt4Dfa4fOEY
>2: https://www.youtube.com/watch?v=WPkoJeVY0xY
>At some point after you've become a celebrated scientist on your home planet, other ayyys make breakthroughs that give them access to more information about human speech and researchers stop taking your classification seriously (though you retain dumb proponents in the public who who straw man: "all human languages are the same"). Eventually an ayy comes along and runs a few audio samples (obtained via audio broadcasts small number of Earth regions) through PCA and by looking only at the first two PCs they conclude that the two clusters can be correlated to "clusters" viewable via the PCs.

In truth we know that there are tons of human languages and their study is incredibly complicated. As a brief overview you have:
>Tons of languages have countless distinct and mutually unintelligible variations (some with hundreds)
>Lots of small and large changes over time
>Relationships between languages are super messy with words and structure being shared in seemingly random ways.
>Can exhibit all sorts of totally different syntactic and semantic features.
>Tons of different meaningful ways to categorize them (that incidentally don't have anything to do with clustering algorithms).

Now, consider that for SNPs the actual dataset has hundreds of millions of dimensions and you are only looking at two PCs generated from a relatively tiny sample size with only about 200,000 dimensions.

I think merging is a huge problem and if we actually did things properly we would find all sorts of complicated but interesting things about SNPs (even without using clustering algorithms). Too bad people are fixated on pseudoscience from the retard ages.

I feel that the whole " race doesn't exist but genetic differences between groups of people exist" is an oxymoron.

Maybe race is not the right word or concept, but doesn't the second statement more or less prove that humans can be classified in a certain way ?

>>9223161
>doesn't the second statement more or less prove that humans can be classified in a certain way ?
Just like your porn can be classified in lots of different ways so can humans. You could classify people based on height, weight, hair color, you name it. The problem comes in when you pic an arbitrary classification based on a mixture of subjective and visual data and then claim it's based on genetics. Then the problem is compounded when you add in a bunch of bad logic and anecdotal evidence to claim that your classification system proves certain things about genetics.

The problem isn't with the concept of groupings. Population Genetics is used to study all sorts of groupings in general.

The problem is in
1) Unrigorous unscientific groupings being used to make logically unsound assertions.
2) Trying to claim that there is only one grouping.

>>9223134
>How is merging not a problem to you?
i didn't say it isn't. rather the point is even after we merge all these clusters there is still meaningful distributions in the data. genomes are divergent geographically on the main axes.

again i'm not using this to say it is all the variation as you have repeatedly strawmanned, just that the biggest variation of the variation is geographically distributed.

>I think merging is a huge problem and if we actually did things properly
right because focusing on the less significant dimensions negates the patterns we see in PC1 and PC2.

>Too bad people are fixated on pseudoscience from the retard ages.
right microevolution is pseudoscience now, ok.


>>9223212
>You could classify people based on height, weight, hair color, you name it.
But genetically people are most different according to ancestry. hence PC1 and PC2 isn't based on height or hair color, etc. if the point is to see which categorization ppl are most genetically different from each other then geographic ancestry is the clear winner here.

>>9223346
> just that the biggest variation of the variation is geographically distributed.
lmao, no you can't even say that when you're ignoring hundreds of millions of SNPs and have a relatively tiny sample size for the true dimension of the data.

>right because focusing on the less significant dimensions negates the patterns we see in PC1 and PC2.
It's just like the analogy with classifying languages using an arbitrary system of how they sound. I guarantee 100% that you could construct a statistical model and/or deep learning classifier that would present a pattern over audio samples putting them into those two fields. At the same time only a fucking retard would think that approach is representative of the data or meaningful in any way.

>microevolution
lmao, only the only people who use that term are anti-evolution brainlets who reject evolution on larger scales. Also, it has nothing to do with race.

>But genetically people are most different according to genetics.
fixed.

>if the point is to see which categorization ppl are most genetically different from each other then geographic ancestry is the clear winner here.
No, that's what population genetics is for. Ancestry is self-reporting social science (i.e. "muh feels say I'm 1,000,000% pure white German").

>>9220754
>however if you see clusters form after a pca, pca may merge
1) There's a difference between PCA and ignoring thousands of dimensions because the first two are at least as significant as each of the ignored dimensions.

2) PCA can distort data including merging and splitting (less common) clusters. For this reason there are many alternatives to PCA (including some that also rely on Singular Value Decomposition) that may give different but semantically similar results.

3) PCA IS NOT A CLUSTERING ALGORITHM, that's not how it's meant to be used. Picking two PCs and visualizing them is really not that different from picking two dimensions in your original dataset and visualizing those (the magic of singular value decomposition only removes redundant data from your set and sorts your dimensions). There are lots of ways to ignore 99.99..% percent of your data, this is one of them.

>>9223402
holy shit you are one butthurt cunt.

>lmao, no you can't even say that when you're ignoring hundreds of millions of SNPs and have a relatively tiny sample size for the true dimension of the data.
in other words you don't know what pca does. PC1 is the axis with most of the variation of all the dimensions. it's not ignoring them, rather finding which axis has most of the variation in the dimensions. yes it can and most likely does in this case will merge a lot of clusters but it will give you back the most variable pattern based on all the data you feed it. it can't split which is the key here.

>At the same time only a fucking retard would think that approach is representative of the data or meaningful in any way.
because you don't understand what pca is.

>lmao, only the only people who use that term are anti-evolution brainlets who reject evolution on larger scales. Also, it has nothing to do with race.
notice how i use the word ancestry here.

>fixed
except the patterns has a close correspondence to geographic ancestry. so close it practically reconstructs the map of europe,

>No, that's what population genetics is for. Ancestry is self-reporting social science (i.e. "muh feels say I'm 1,000,000% pure white German").
and part of pop gen is looking for patterns in gene flow. geography plays a role in evolution and has played a role in man's evolution. sorry reality doesn't conform to your 1970s preconceptions about human genetics. sorry but looking at whole genomes africans are more dissimilar to europeans than they are to themselves in almost every case. see: >>9218352
and that's not "ignoring" dimensions.

anyway i'm done here i have work to do.

>>9223404
tell all that to: >>9223402

i never said pca ignores data, he repeatedly has. so fucking desperate like a /poltard scurrying to deny his african ancestry results. anyway i'm out.

>>9223404
just 1 thing, p sure it doesn't split tho.

>>9223423
>being this illiterate.

>lul I'm going to ignore data by only looking at 2 PCs and ignoring the rest.
>>anon don't you know you could be missing out by not ignoring data?
>wot tha fuk are you talking about? I'm using PCA and PCA doesn't ignore data!!!!!

kys yourself, retard.

>>9223432
>>anon don't you know you could be missing out by ignoring data?
whoops, typo

>>9223432
>lul I'm going to ignore data by only looking at 2 PCs and ignoring the rest.
and how are the pcs obtained? you do realize a pc does go through the dimensions in most cases, right? so pc1 is going through thousands of dimensions in this case.

you are the retard, retard. goodbye.

>>9223465
Singular value decomposition doesn't just fucking magically capture information from every dimension, retard. Learn some fucking math or get off /sci/.

>>9223477
>i'm just gonna strawman because that's all i have
k, bye. idk why i even bothered arguing with you. you were disingenuous from the start.

>>9223485
>anon, the math doesn't work that way.
>that's a strawman!!

lmao, maybe you should've paid attention in linear algebra, anon.

>>9223501
i didn't read this >>9223404 when i responded to it, just saw the words "PCA and ignoring thousands of dimensions" and assumed it was the other anon, but it shows the depths you don't get what i'm getting at. pca doesn't ignore dimensions, pc often go through them and ranks the pcs according to variance. and as i have routinely stated if we are looking for visualizing how most of the variance looks like this is perfectly valid. and it so happens ppl are most varied ancestrally. so much so the first 2 pcs recreate geographical maps.

anyway i'm out. i kno youw want your 1970s fantasies to be reality so bad like pol doesn't want to have a black granddad but i'm sorry this is thw world we live in. deal with it. and bye.

>>9223513
Read:
https://en.wikipedia.org/wiki/Principal_component_analysis#Details

Suppose you found some data behind a dumpster; it has 3 dimensions and seems to be about local houses. The columns in your data are:
>Internet connection speed
>House size in square feet
>House size in square meters
Now, suppose you were to run PCA on this. It would most likely two PCs, one for
>Internet connection speed
and one for
>House size
Since internet connection speed and house size are independent variables and PCA is a linear transformation then it won't magically merge them into a single PC. Also, due to measurement discrepancies (i.e. maybe two different people measured the house sizes independently leading to slightly different values) the new PC for house size won't be exactly the same as the other two but it will try to capture as much data.

Now, suppose some fucking dipshit comes along and decides that only the first PC is necessary because "lul it captures basically all of the variance rite?". That dipshit is basically you.

>>9223513
>it was the other anon,
Also, they're both me. You're just fucking illiterate.

>>9223575
>Internet connection speed
>House size in square feet
>House size in square meters
amazing example senpai, notice how the dimensions aren't even of the same metric. unlike with distance measures of 4 base sequences. i agree here pca is a horrible approach. but not when trying to see variances in genomic data.

>Now, suppose some fucking dipshit comes along and decides that only the first PC is necessary because "lul it captures basically all of the variance rite?".
you just keep proving my point, you're a disingenuous fuck.

>>9223577
again what did i say you illiterate dipshit?
here:
>i didn't read this >>9223404 when i responded to it, just saw the words "PCA and ignoring thousands of dimensions" and assumed it was the other anon
>assumed
because i didnt read all your gibbering.

you are such a blatantly dishonest person. so desperate to have your 1970s fantasies a reality.

>>9223594
>amazing example senpai, notice how the dimensions aren't even of the same metric. unlike with distance measures of 4 base sequences.
It does not make one bit of difference whether or not they're the same metric. What matters is data independence.

>i agree here pca is a horrible approach. but not when trying to see variances in genomic data.
PCA itself isn't bad, ignoring 99.99% of the PCs it produces is bad.
^^^^READ THIS LINE 100 TIMES because I guaran-fucking-tee you will fuck up at reading once again.

The point is that all of the dimensions that PCA produces are linearly independent and thus it is 100% wrong to say that the first two PCs tell you something about the other 99.99% of the PCs you're ignoring.

>>9223594
>I'm not illiterate! I just didn't read!!
Excellent argument, anon. By the way, both of these are me too >>9223423

>>9223620
>PCA itself isn't bad, ignoring 99.99% of the PCs it produces is bad.
well that depends what if the 2 first pcs do actually capture 99% of the variation?

but i misspoke too looking back. i see where our misunderstanding is from i said most when i should've said largest variation for pc1.

but back to my point if the point is to see how the largest variation in sequences looks 2d if this is stressed there is nothing invalid. and you gotta admit it's neat how the first 2 pcs correspond to geography like that.

>Excellent argument, anon. By the way, both of these are me too >>9223423 (You)
no you see this entire thread has been me arguing with myself when i should be doing work.

>>9220441
>In school we measured bone lengths and structure and we found that the bone lengths differed in the humerus bones of asian people, white people, and black people. Visually shapes of the skulls are anatomically different. There is a physical difference and you can tell which skull is which once you learn the characteristics. Though there is a slight percentage you can get a race wrong by just using one technique.
What about multi-ethnic people

>>9206519
While we are self-conscious and advanced compared to most other animals, we are still animals and at the mercy (to a point) of our genes, and these genes tend to cluster and produce higher frequencies of certain traits in what could be called different races.

In the case of completely global random admixture, there would truly be one race with very loose trends in genotype/phenotype, but this is not and has never been the case, ultimately mountains and oceans and other physical barriers will isolate certain genetic aspects and force selection in favor of locally preferred traits.

>>9223962
>Researchers focusing solely on whether their results are statistically significant might report findings that are not substantive[37] and not replicable.[38][39] There is also a difference between statistical significance and practical significance. A study that is found to be statistically significant, may not necessarily be practically significant.[40]
gtfo brainlet

>>9206366
Funny how it is only /pol/ threads that get more than 100 replies.
/sci/ is a crappy /pol/ colony due to the tiny size of this board.

>>9224016
You barely see any /pol/ replies in this thread.