Paweł Lenartowicz @plenartowicz@qoto.org

@david_colquhoun @lakens

I'm quite sure that I agree with you. But, in my opinion, it's problem with lack of rigour and putting too much confidence in "statistical ritual" than p-value itself. Let me provide some example:

1) We are using p-value to detect candidate genes in some traits, like depression, anxiety, intelligence...

2) We can conduct some meta-analysis of such analysis, to make sure that candidate gene has effect, and obtain some "satisfying low p-value", like p = .00002 (σ > 4) and based on these results to conclude that we have strong evidence for effect.

Point "2)" is wrong, and there is a little evidence in it, but I think, that using p-value to detect "candidates" is defendable. Or, at least, I'm not aware of any better method.

@david_colquhoun @lakens

Fisherian aproach was intoduced in 1925, while Jerzy Neyman proposed his solution in 1926.
It's amazing that after almost 100 years, there are lively discussions about them!

In my opinion, these are two different, not always competitive tools, like screws and nails.
If we have specified hypothesis to test, it'll be better to use Neyman approach and calculate LR or other suitable test, with satisfying α and β.

But, when we want to just test null hypothesis for error detection, or basic exploratory analysis, p-value seems to be good approach.

@david_colquhoun @raoulvanoosten @lakens

Except Schrodinger's cat ;)

@OskarA
It's amazing that such ignorant and genocide denier as Chomsky is still considered an expert.

@psychNerdJae @ct_bergstrom

You can scatterplot everything, but that doesn't mean it makes sense. Especially plotting ranked data doesn't. In the best scenario, it's waste of time, for both authors and readers.

Also, linear regression is the wrong method to analyze such data. None of the assumptions were met!

But, IMO, the problem is field-wide: correlation study instead of an experiment, no data analysis plan at preregistration, no power analysis, non-random sample, no clear relation between theory and operationalization, big potential for p-hacking and so on...

@ct_bergstrom
"Evolutionary psychology" is largely a pseudscience.

@uebernerd
It's clearly statistics with p-value:
1:1 => p=0.3173105188
2:1 => p=0.0455001240
3:1 => p=0.0026999344
4:1 => p=0.0000633721
5:1 => p=0.0000005742

But the problem is elsewhere. In experimental social science "effect size" and "sampling error" are artifacts of method, when data rarely meet the assumptions.

Triangulation is IMHO better idea than efforts to lower p-values: https://www.nature.com/articles/d41586-018-01023-3

@atomicpoet @martinvermeer
Maybe, you would like something like graphene OS or LineageOS, which are non-google Android distros.

@atomicpoet @martinvermeer

There is plenty of alternative OS. And plenty of Android distribution too: https://en.wikipedia.org/wiki/List_of_custom_Android_distributions

@mgaruccio @atomicpoet
Also, isn't Apple using Darwin, which is UNIX(like) Kernel?

@lakens
What a beautiful ergodic distribution <3

@uebernerd
Yes, but:
1) there should be some robust system for identifying and archiving articles (in case of a blog-site shutdown, or similar);
2) In my opinion, the peer-review system is ineffective and biased, however, I think that without it, process would be much more (status) biased and vulnerable to fraud, corruption, conflict of interests and so on.

@breadandcircuses @Caitoz
I'm disgusted with the manipulation of the authors of this graph.
They are comparing the mean to the median, of, even not every worker, but some part of them, "traditional blue collars"
which decrease since 70'...
More than half of this "disconnection" is clearly a lie, the second part is about growing inequality. Which is catastrophically bad, but we do not have to lie to show or prove that.