/sci/ - Science & Math

[Anonymous No.11194800]

Hello, stats question please. Confidence intervals.

Let's say I have 5,000 daily values in a time series, and that the sample standard deviation of those 5,000 values is 10, with a mean of, say, 0.

I would say that, with 90% confidence, the next day's value will be between -10 and 10, right?

But I had this discussion with someone who says that you need to divide the SD by sqrt(5000-1), -> (-.14, .14)

I of course know the formula that they are using, but I forget why and when that's used, seems retarded here. What's going on? pic is dog

[El Arcón No.11194807]

>>11194800
>What?
The Israelites forsook the Lord and turned to the whore of Babylon instead.

[Anonymous No.11194811]

>>11194807
Yes I know that now please let's discuss something that isn't completely obvious.

[El Arcón No.11194842]

>>11194811
I think mathematicians should entertain the notion that the point density of abstract mathematical points is infinity points per point rather than the one point per point which is commonly assumed. What is your opinion?

[Solivagus No.11194846]

>>11194842
Our opinion is that infinity is the shared point of reflection for all algebraic geometry, or word salad potato monkey.

[El Arcón No.11194855]

>>11194846
I will take away your blessings.

[Solivagus No.11194859]

>>11194855
What blessings? I'm Satan. I don't have any blessings. Pretty sure you can't demote me to anything less than Satan at this point, God, unless you don't want to have someone to put in your magical 'timeout' room.

[Anonymous No.11194861]

You're both wrong, but your friend is closer. When determining the margin of error, you divide the standard deviation by the square root of the sample size, not the sample size -1. Also, you multiply that by the z-score that corresponds to your give confidence interval. For 90%, that's 1.645. Then you add and subtract that product from the mean.

[Anonymous No.11194862]

>>11194800
If the values are normally-distributed, then the 90% confidence interval for the next value will be ±10*Z_0.95 = ( -16.45, +16.45)

[El Arcón No.11194866]

>>11194859
If there weren't things in the world that you care about, things that I can take away from you and use to hurt you with, then you wouldn't be in these threads. If you had nothing, you wouldn't care.

[Solivagus No.11194883]

>>11194866
Why would that be the conclusion I reach, you fucking retarded American mongoloid?

[El Arcón No.11194891]

>>11194883
Just like me, you make those posts you make because you want someone to like what you write.

[Solivagus No.11194893]

>>11194866
>>11194883
>Watch how Tooker's rage boner increases 500%!

[Anonymous No.11194899]

>>11194807
>>11194842
>>11194846
>>11194855
>>11194859
>>11194866
>>11194883
>>11194891
>>11194893
i'm 12 whats happening

[Anonymous No.11194904]

>>11194899
>whats happening
two retards decided to have a pissing contest in a statistics thread

[Solivagus No.11194905]

>>11194891
Huh? Wow, how much self-esteem do you NOT have there Tooker? Serious question.

[Solivagus No.11194907]

>>11194904
Statistically speaking, all threads are Tooker threads.

[El Arcón No.11194920]

>>11194905
Even if you are ready to pay the price I will extract from you, your loved ones certainly are not ready to pay it in kind.

[Anonymous No.11194923]

>>11194862
oh shit is this true? am i retard?

[Anonymous No.11194925]

>>11194893
The Sicilian kike!
>I
>N
>C
>O
>N
>C
>E
>I
>V
>A
>B
>L
>E
>!

[Anonymous No.11194926]

>>11194923
you just weren't paying attention

[Anonymous No.11194951]

>>11194862
>>11194923
>>11194926
Its (-.233, .233).

[Anonymous No.11194961]

>>11194951
>>11194926
well for real though, what is the interpretation of the mean +/- z*std/(sqrt(n-1)) when we are looking at drawing one additional datapoint? Maybe the interpretation is something like, if we draw "n" additional datapoints, then their average would be within that interval 90% of the time or whatever. But if the true standard deviation is 10 (mean 0) then I feel like, well okay yeah for 90%, z*10 -> (-16.45, 16.45) makes sense. But I don't see when the fuck you would divide the SD by 5000 or whatever. However that's a mainstream formula...

[Solivagus No.11194966]

>>11194920
Since when was I loved? This is news to me. May I know who this person(s) is/are?

[Anonymous No.11194967]

>>11194899
a literal schizophrenic argues with the LORD of /sci/

[Solivagus No.11194971]

>>11194967
A literal autistic sociopathic schizophrenic argues with the magic cave wall that moves and shimmers as the rain falls washes away all the paintings I, Ug the caveman, made.

[Anonymous No.11194980]

>>11194961
Okay I get it now. Thank you apples you are my greatest ally. Basically i'm right. Yeah, with 5000 observations, and a sample mean of zero and a sd of 10, yeah we would estimate the true mean to be like ( >>11194951 ) what you said with 90% confidence.

Okay that makes sense.

But what I am talking about is, if we are to draw one more piece of data, what is a 90% CI of its expected value? Okay and that is -16,16? That seems reasonable too.

Okay pack in up I think we're done thanks for entertain.

[Anonymous No.11194981]

>>11194961
Where are you even getting that first form of it? You always divide by the square root of the sample size, because using a larger sample size allows you to be more accurate. Imagine using a sample size of 3: your interval would become (-9.527, 9.527). It's still a 90% confidence interval of a normally distributed population with a standard deviation of 10, but because the sample is so much smaller, the interval is wider, meaning less precision.

[Anonymous No.11194984]

>>11194980
No, you would use the same formula but with a sample size of 5001.

[Anonymous No.11194985]

>>11194800
>reddit spacing
paypal me or find me on Chegg - Phil F

[Anonymous No.11195009]

>>11194981
>>11194981
The sample standard deviation is already divided by the square root of the sample size.

[Solivagus No.11195037]

>>11195009
What is a square root supposed to indicate from the logic of perspective? Does irrational = shared?

[Anonymous No.11195046]

>>11195009
No it’s not. It’s the square root of the sum of the squares over the sample size.

[Anonymous No.11195060]

>>11195046
same thing tho unironically?

[Anonymous No.11195810]

>>11195060
It helps if you know what standard deviation represents, as well as how to calculate it. You find the difference between each datum and the mean, square the difference to make them all positive, then divide by the sample size. This gives you the variance, which represents to total variation of your sample. Taking the square root of the variance gives you the standard deviation, which is the average difference any datum is from the mean.

Going back to confidence intervals, you further divide the standard deviation by the square root of the sample size before multiplying with the z-score. Why? Because you can arrive at the same standard deviation with two samples of wildly different sizes. The larger your sample size, the more precise you can be in determining the 90% CI. See >>11194981
>Imagine using a sample size of 3: your interval would become (-9.527, 9.527). It's still a 90% confidence interval of a normally distributed population with a standard deviation of 10, but because the sample is so much smaller, the interval is wider, meaning less precision.

[El Arcón No.11195827]

[Anonymous No.11197135]

>>11194800
>>11194862
>>11194923
>>11194926
>>11194961
>>11194980
>>11195009
>>11195060
You should have to pass a test to post here.

[El Arcón No.11198179]

I wonder at times if the Strzok "insurance policy" was that Exide inserted into my 2016 hiring paperwork fraud papers saying that I have authorized them to drug/rape/hypnotize/censor/torture/kidnap/mutilate me as much as they want, or that the job I was being offered was something other than the lowest ranking SQL bitch in the IT department of manufacturer and vendor of electrical batteries. Although all of the conditions of fraudulent misrepresentation are plainly met:
>https://smallbusiness.findlaw.com/business-laws-and-regulations/fraudulent-misrepresentation.html
>The Elements of Fraudulent Misrepresentation
>In order to prevail in a lawsuit for fraudulent misrepresentation, the plaintiff must be able to prove the following six elements:
>A representation was made (in contract law, a representation is any action or conduct that can be turned into a statement of fact)
>The representation was false
>The representation, when made, was either known to be false or made recklessly without knowledge of its truth
>The representation was made with the intention that the other party rely on it
>The other party did, in fact, rely on the representation.
>The other party suffered damages as a result of relying on the representation
I would not be surprised at all to learn the the USA courts have determined that Exide's fraud contract is valid because their conspiracy to defraud me did bear the fruit of my signature on their fraud contracts when I signed them in good faith after they were represented to me as being related to the terms of employment I discussed. Furthermore, I would not be surprised in the least to learn that Chief Justice John Roberts himself was the one who made to me these fraudulent misrepresentations when he told me the hiring paperwork was related to the terms discussed: 40hrs of SQL-themed bitch work for an electrical battery manufacturer and absolutely nothing to do with me authorizing them to drug/rape/hypnotize/censor/torture/kidnap/mutilate me.

[El Arcón No.11199693]

http://s000.tinyupload.com/index.php?file_id=53054654562431378856