>p-values
>women
>>8523169
seconded.
>>8523169
amen, don't need any of that.
>>8523169
>tfw 50% of the population has an IQ below 100.
>>8523167
yeah thats why they stick to bio, psychology and make up
>>8523173
>tfw 50% of the population has an IQ below 100.
das rasis
>>8523177
sexist you mean. Lol i know for a fact, australia doesnt have a shit iq.
>>8523169
women not intelligent enough to do science so i agree, no place.
>>8523167
>women
>>8523167
>REAL science
Sophomoric
>>8523182
>australia doesnt have a shit iq
You haven't spent enough time on boards with flags
>>8523223
u should direct me?
>>8523167
who would you choose, tim hunt or women?
>>8523167
>Bayesian priors
p-values are actually of value in the physical sciences b/c the underlying distribution is more likely known.
In the social sciences they are almost always bullshit.
>>8523240
The point is that they shouldn't be doing it period. The results they obtain are not statistically valid.
What they are doing is called scientism it is a cancer.
>>8523243
why is this? i hear about it but dont know anything about it.
>>8523248
You have to assume an underlying distribution when you calculate a p-value
DARK:
>energy
>matter
>thrust
>photons
>anal cavities
>>8523255
oh fair enough. wjat else would people do though.
>>8523265
That's the million dollar question. The point is that p-values and frequentist statistics more generally were invented to address specific situations where the underlying distribution could be known to a reasonable degree of certainty. If social scientists want their field to have the same level of respectability that the hard sciences currently enjoy, the way to do that is to invent new techniques appropriate for their field of study, not to mindlessly copy the technical language of a more respectable field.
>>8523277
well social scientists and psychologists usually don't know shit about statistics so i guess its fucked. what about bayesian testing.
>>8523285
I don't see how it addresses the concerns in my previous post.
Imaginary numbers.
>>8523321
of the data-generating process
Paradoxes, any of them.
>>8523321
central limit theorem doesn't actually represent the distribution of the population necessarily though does it? and sample sizes too small?
>>8523306
what about guessing. just "this number bigger this"
>>8523328
>>8523339
so the results then
you don't have to assume anything about the underlying distribution and depending on the number of variables you have to control for, it could remain unknown for all you care. The point of statistics is that for enough tests, the results will converge to a normal distribution and this belief has a firm mathematical basis. What you're studying may or may not follow a standard distribution, but your guesses will. That is why statistics work.
>>8523357
Social "Scientist" detected lol
full retard
>>8523357
No
the central limit theorem implies that the distribution of the (suitably normalized) population mean is asymptotically normal, but gives no information about other statistics
for example, suppose you flip a coin repeatedly and are interested in the first time you get a heads. The central limit theorem says nothing about the distribution of this, even in the limit of many tosses.
>>8523182
That image actually shows the average IQ of the native populations of the respective countries. So, yes the average IQ of an abo is ~60, but they're barely human.
>>8523169
>female """psychologists"""
>>8523452
What's your point? you're not saying anything new here. Again, no one doing statistics assumes the underlying distribution is a normal one. In fact, it's a safer bet to say most things you'll study won't have a normal distribution. The point, again, is that it doesn't matter because "guessing" follows a normal distribution.
>The central limit theorem says nothing about the distribution of this
what are you talking about It couldn't express more clearly that the probability of being right given you guess heads on a certain toss number would follow that of a normal distribution after normalization. I feel like you don't even understand the points you're making.
>>8523514
Actually, I feel like you're the one who doesn't understand the points I'm making.
>no one doing statistics assumes the underlying distribution is a normal one
In fact, many common statistical tests are based on this assumption, for example the student's t-test.
>what are you talking about
the clt describes the asymptotic distribution of
[math] (\sum_{i=1}^n X_i)/\sqrt{n} [/math] where X_i are centered iid random variables.
The statistic I gave is of the form [math] \inf \{n: X_n=H\} [/math]. I would be very interested to see how the distribution of this could be inferred from the clt.
> the probability of being right given you guess heads on a certain toss number
this is not the same as the probability of the given toss being the first head thus far
>>8523571
>I would be very interested to see how the distribution of this could be inferred from the clt
this is exactly why i'm saying I don't think you understand the points you're making. Yet again, I never said you could infer any distribution from anything. I said twice that it doesn't matter. This feels similar to an XY problem. What is the point of making this point? how do you think this invalidates the use of p-values.
>this is not the same as the probability of the given toss being the first head thus far
Sorry, I tried to rephrase your question in a way that makes sense. If I interpreted it the way i think you're literally saying it, it feels like gambler's fallacy tier stuff which statistics obviously can't answer.
>>8523618
> Yet again, I never said you could infer any distribution from anything. I said twice that it doesn't matter...how do you think this invalidates the use of p-values
because the underlying distribution appears in the definition of a p-value. so you can't calculate a p-value unless you know (or assume) the underlying distribution
>gambler's fallacy tier stuff which statistics obviously can't answer
I'm not sure what this is supposed to mean. It's a well-defined random variable with a known distribution.
>>8523182
Account for the natives who ooga booga no fire
>>8523167
>egos
>>8523167
> not understanding basic statistics
sounds like a shit scientist mate