p-value

[Anonymous No.15693851]

So you have a real life process.
This real life process produces measurable outputs.
You collect these measurements.
You notice the measurements appear to contain some random variation.
You make some assumptions about the process and produce an ideal statistical model that takes the random variation into account.
Now you make a hypothesis about one of the parameters of this model. That is, you assume that the parameter takes some value.
And then, using trickery from statistics, you devise how to calculate some test statistic from your data, and then if this value has a sufficiently low probability of occuring (either in a frequentist i do the experiment again sense, or in a bayesian what sort of confidence do I have given these observations) you reject that the parameter has this value?

Did I get that right?

[Anonymous No.15693856]

What a faggot this guy turned out to be, who could possibly have guessed that the broadspectrum "conservative" media promotion engine could possibly be this way.

[Anonymous No.15693862]

>>15693851
You're still losing your mind with rage from yesterday over having your determinitard religion refuted. Did I get that right? :^)

[Anonymous No.15693871]

>>15693851
>vague pontification about meaningless hypotheticals with no concrete value
Ask a real question about something that exists and I'll consider answering.

[Anonymous No.15693874]

I should have prefaced my thread with a triple digit IQ requirement.

[Anonymous No.15693883]

>>15693874
Good thing you didn't, otherwise you wouldn't be able to post in it and that would be awfully embarrassing.

[Anonymous No.15693906]

>>15693883
if you can't explain you can just admit you know

[Anonymous No.15693909]

>>15693906
Ask a real question first.

[Anonymous No.15693926]

>>15693909
What makes the question fake?

[Anonymous No.15693934]

>>15693926
The fact that it's 100% hypothetical and involves only manipulable placeholders that can be used to move goalposts and generally engage in pilpul tactics. Ask a concrete question about things that exist in reality or there's no point.

[Anonymous No.15693940]

>>15693934
NTA but it sounds like you're just a nigger incapable of abstract thought.

[Anonymous No.15694030]

I have to agree with this anon>>15693940
One of the hallmarks of intelligence is recognizing concepts outside the textbook.

[Cult of Passion No.15694059]

>>15694030
>I passed the test you have to do what I say now.

[Anonymous No.15694072]

You assume a real-life phenomenon follows some DGP. If this DGP is true you have some expectations about its behavior. So you devise a statistic to test this behavior. If the DGP is true, the number should be close enough to what you would expect. The further it is from expectation and the less likely this outcome under your hypothesis the more suspicious you should be that your DGP assumption is incorrect.

[Anonymous No.15694099]

>>15694072
Right, but how do you select an ideal DGP to get the best possible test statistic?

[Anonymous No.15694335]

>>15693851
The way i see it is the distributions that underlie the data you are gathering is what your re attempting to determine. You are using the p-value as a metric for determining which of the distributions are most likely to explain the data points gathered thus far.

To select the appropriate distribution there are a couple methods. Logic is the first one, each distribution has a fundamental method by which it is generated. Identifying how these fundamental methods relate to the process at hand gives us insight into which distribution is appropriate. The other is to basically just fit all distributions you can computationally and compare how well their associated test-statistics support the assumption that the distribution represents the data.

>>15693871
>>15693934
>>15694059
The fuck is this tard doing replying to anyone here.

[Anonymous No.15694343]

>>15694335
I can jive with this. What surprised me with statistics is that if you just look at the data in various ways, graph it and perform transformations and so on, it often becomes much more clearer what its properties are, rather than just computing some statistics and talking out your ass

[Anonymous No.15694345]

>>15694335
>not smart enough to recognize a change in posters
>even when one of them namefags

[Anonymous No.15694353]

>>15694343
Yeah, when I'm analyzing new data I always start with visualization of some representative subsets. Computing random statistics at it risks over fitting super hard. This is the curse of ML cause a lot of people just toss ML techniques at their data instead of attempting to understand what their data is and over fit their models.

>>15694345
God your easy to bait

[Anonymous No.15694362]

>>15694353
My easy to bait what? Finish your sentence.

[Anonymous No.15694408]

>>15694362
Trolling is a art.

Lurk more

[Anonymous No.15694410]

>>15694408
*an