Why should tax money be used to finance pseudosciences like psychology and sociology...

Why should tax money be used to finance pseudosciences like psychology and sociology? Most of them don't even understand stats 101 (e.g. what is a p-value).

>at least one mistake

That means each one could have potentially made just one mistake, which is good.

No it's not when the questions are extremely simple questions that can be answered true/false.

Can Sup Forums get this right?

>german psychologists
>german
That's your problem.

Anything where the sample size is less than 30 is complete and utter bullshit

Oh, wait, t-test. Sorry

I can, but it's retarded because the first pic says they're all wrong.

Also, everyone gets this wrong. Even a lot of statisticians will.

Now how about you, OP - can you define what a 95% confidence interval means?

Let's say the value 5.6 is included/not included in my 95% confidence interval. What does that say about 5.6?

This is a test, and I will laugh at you if you get it wrong.

That's a really dumb statement.

The 1st LIGO study had a size of 2 (samples of 1 event).

>even a lot of statisticians get it wrong
only if they are german statisticians

Can you pass my test though? Cause if you can't, it's you who's bad at stats.

It's a dumb enough statement I believe a German with a proxy said it

Read my second post, right after the one you replied to. Didn't realize it was a student distribution, because I'm a lazy slavshit

Still fucking stupid.

I have a vague idea of what you're going for, but what you verbatim said was dumb.

I'm gonna bet you're statistically illiterate.

Not OP

A confidence interval is, given a distribution, an interval that captures, say if we have a 95% CI, 95% of the samples taken from that distribution.

It's simple enough even a few Germans may get it.

>Germans calling someone dumb

Okay, you can cite a definition (in a way that's sufficiently vague as to not be wrong), but I didn't ask you to cite a definition.

Your reading comprehension is bad. I didn't call anyone bad. I called one person statistically illiterate, and one idea bad.

I meant that that for a sample to be representative, it has to be bigger than 30 so that it can be closer to a normal distribution, instead of a flawed/skewered t-distribution, all other things being the same.
I was just lqzy and didn't read through the test

What's vague about that definition?

Anyway, if 5.6 is outside your 95% CI, that simply means, given that CI is defined on a distribution that represents your null hypothesis, a value that would occur less than 5% of the time given your null hypothesis is true.

That's still fucking wrong though.
Like, really really wrong. You could quantify it with something like "in a Z test, ...", but 1. you aren't, 2. nobody uses Z tests, especially not with small samples.

>a value that would occur less than 5% of the time given your null hypothesis is true.
I have no idea what this means, but I was asking what it says about the particular value 5.6.

>a value that would occur less than 5% of the time given your null hypothesis is true.
A frequentist null hypothesis is typically not "represented by a distribution", but concerns a point hypothesis.

Hint: you're missing the trick in the question. Yes, it's a trick test, and you're currently not passing.

this doesn't look like a distribution to you?

The blue PDF is not the null hypothesis. In frequentism, the parameter isn't a random variables. (Test statistics are, and thus have a distribution.)
You're thinking Bayes.
(Maybe. I'm not sure what you're thinking. I can try explaining it to you if you can state a bit more precisely what you understand a null hypothesis to be?)

Of course the parameter isn't random here. The distribution represents samples (point estimates) of a fixed parameter of interest.

You can shift that distribution (or shift it's mean more precisely) so it is around the null and check if your estimate is within that interval, or have the distribution centered at your estimate and check if your null point is within that CI. It makes no difference.

>samples (point estimates) of a fixed parameter of interest
This is again confused. Samples are drawn FROM the real distribution. Samples don't REPRESENT the distribution. And they're clearly not samples of the parameter.

>You can shift that distribution (or shift it's mean more precisely)
If you shift the mean, you shift the distribution et vice versa.

>it is around the null and check if your estimate is within that interval, or have the distribution centered at your estimate and check if your null point is within that CI. It makes no difference.
Oh god. Who taught you statistics? I'm not trying to be insulting, but you're really confused about everything.

Of particular note:
>have the distribution centered at your estimate and check if your null point is within that CI
The CI is not a distribution either.

You must really get clear on what these mean:
>parameter
>distribution
>hypothesis
>random variable
>sample

Because you're getting all of them wrong.

I'm not going to be typing out long ass explanations here where you nitpick details of things I haven't given af about for years (that is, retarded approaches to inference). You fully well understand what I mean.

If your 95% CI is say [5.4, 5.5], then your point estimate 5.6 is extreme enough to occur less than 5% of the time given that we drew that sample from the same distribution the CI is defined on.

Say it is symmetrical, and the "null mean" is 5.45, then we can "shift" this so that it is around 5.6, i.e. [5.55, 5.65]. Now, our "null mean" instead is outside the interval, like the estimate was outside the earlier interval.

Besides it's past midnight and I'm drunk.

>Besides it's past midnight and I'm drunk.
It shows. What you're saying doesn't make any sense.

And pointing out that parameters don't have distributions (=> instead, distributions have parameters), or that parameters aren't random variables, isn't a nitpick.

> your point estimate 5.6 is extreme enough to occur less than 5% of the time given that we drew that sample from the same distribution the CI is defined on.
blah blah blah

This truly makes no sense. You're using some of the correct phrases, but they're not put together in any meaningful way.

It would be correct to say that the CI contains the range of parameters for which, were they the true parameter, a test statistic deviating as strongly or stronger from the parameter as that of the data occurs more often than in 5% of samples from a distribution characterized by that parameter.

I didn't state parameters have distributions. Samples (point estimates of that parameter) will fall on a distribution though, but perhaps I was hasty with some statement somewhere.

Now, your last statement is more or less

>Samples (point estimates of that parameter)
No no no no no no no.

Samples are samples. Test statistics (functions of samples) can be used to construct point estimates. And they don't "fall on a distribution", they have a distribution.

This may seem trivial nitpicks to you, but it's really crucial, and if you understood even the basic gist of this, you'd understand why my description and yours are far from equivalent.

>I didn't state parameters have distributions
You said
>a distribution that represents your null hypothesis
No. A null hypothesis concerns a parameter.
This isn't trivial. This is the absolute basics, and you're extremely confused about them.

I hope what you take from this is some modesty.

I was hasty with some of my statements and you decided to sperg out

>I was hasty with some of my statements
The point is not that your sentences were confused, but that your thoughts are confused. You don't understand statistics.

So will you take some modesty and perhaps curiosity from this, or go "I TROL U"?

Not really. I don't see any misunderstandings, just a lack of precise use of some terms.

>I don't see any misunderstandings
That - that you don't see it - is the point.

You understand literally nothing from statistics. You confuse basic terms. Better listen to german and polish mathematicians and derivatives, we know better.

We fund these piles of wank because women take these subjects, and God forbid we stand in the way of a woman being "educated".

Because they've been used for years to understand, control and manipulate people through advertisements, news and politics. Few good documentaries about this are 'the shock doctrine' and 'century of us.'