🧵 Statistical significance is the biggest scam ever
Anonymous at Sun, 27 Oct 2024 12:23:31 UTC No. 16452019
All those statistically insignificant results are in the trash as a result.
Anonymous at Sun, 27 Oct 2024 12:36:51 UTC No. 16452033
there's a gerd gigerenzer quote about how he'd found a paper where they'd seemingly carried out the statistical significance ritual on their page numbers...
He calls them meaningless statistics.
It's just one more piece of evidence that science has been morphed into the scientism religion
Anonymous at Sun, 27 Oct 2024 12:38:23 UTC No. 16452035
>>16452033
Good talk if you're interested.
https://www.youtube.com/watch?v=er0
Anonymous at Sun, 27 Oct 2024 13:34:06 UTC No. 16452122
>>16452033
>It's just one more piece of evidence that science has been morphed into the scientism religion
Agreed
Anonymous at Sun, 27 Oct 2024 13:34:28 UTC No. 16452125
>>16452019
It's because soft sciences started using statistical significance in their analysis which were not models, made no predictions and had no way to be falsifiable. In other words they were randomly (no pun intended) slapping p-values on patterns from likert scales with no way to falsify them so they're pointless
Anonymous at Sun, 27 Oct 2024 13:47:00 UTC No. 16452155
>>16452019
>All those statistically insignificant results are in the trash as a result.
And the opposite is also true, many statistically significant results from underpowered studies are considered when they shouldn't.
Anonymous at Sun, 27 Oct 2024 17:22:10 UTC No. 16452434
>>16452019
correct.
https://afiodorov.github.io/2015/06
Anonymous at Mon, 28 Oct 2024 01:08:43 UTC No. 16453031
>>16452019
statistics isnt science.
Anonymous at Tue, 29 Oct 2024 15:04:28 UTC No. 16454805
not peer review or consensus science?
Anonymous at Tue, 29 Oct 2024 16:36:24 UTC No. 16454890
>>16453031
Statistics is fine and most statisticians are actually skeptical of how stats are applied. Statisticians are based as fuck and a lot of them are philosophically informed about stuff like philosophy of science and the replication crisis. Applied scientists, and in particular, economists, psychologists, and other social scientists give statistics a bad name because they apply statistical techniques in a haphazard manner that basically yields an pseudo-scientific rationalization of your pre-existing biases and views in a manner that gives a false impression of rigor and objectivity.
Anonymous at Tue, 29 Oct 2024 17:42:48 UTC No. 16454976
>>16452035
>https://www.youtube.com/watch?v=er
good talk but why the fuck does this euro guy have tds in march 2024?
>"they want to save science, at the same time they run the danger that maybe Donald Trump, someone else, will use this to totally cut funding."
Anonymous at Tue, 29 Oct 2024 17:46:49 UTC No. 16454983
>>16452019
All the stuff we are trying to get
Anonymous at Tue, 29 Oct 2024 17:49:52 UTC No. 16454989
>>16454976
I dont think he does. He was just referncing an attitude people might have. He also mentions later in the talk that part of the motive for our flawed approach to p-values was that academics and publishers were obsessed with muh Russia and muh communism, and Jerzy Neyman was from Russia, so the academic community just ignored his work. Not to mention, Karl Pearson and Egon Pearson were chuds obsessed muh Germanic culture and muh genetics.
Anonymous at Tue, 29 Oct 2024 17:53:18 UTC No. 16454991
>>16454989
Etc etc, antoine saint exupery knew this
They can't see facts they see people
Anonymous at Wed, 30 Oct 2024 20:45:34 UTC No. 16456232
most people don't trust stats or evidence they trust people telling them the stats are good or bad
otherwise we don't actively trust evidence in front of us unless we are forced to authenticate it ourselves
Anonymous at Wed, 30 Oct 2024 21:51:41 UTC No. 16456299
>It's just one more piece of evidence that science has been morphed into the scientism religion
Agree, there is no new immaculate science being made, true sceintifc advances are hidden from us, muh science its just big corp shilling for their products or ideology
Anonymous at Wed, 30 Oct 2024 22:04:38 UTC No. 16456316
Look, everyone! 99% of people report their ADHD amphetamines after 6 months to be beneficial! Highest rated appreciation of treatment was in those restricted to staged supply.
Anonymous at Wed, 30 Oct 2024 22:42:11 UTC No. 16456358
>>16452125
Isn't randomly finding significant correlations in a pile of data a clue that there might be a valid hypothesis to be found? I understand that writing the conclusion first is fraudulent but it's a valid input for the start of a study no?
Anonymous at Wed, 30 Oct 2024 23:31:24 UTC No. 16456390
>>16456358
no because by definition you'll find "significant correlations" in 1 out of every 20 things you look at (for p<0.05), even if your data is all completely random. If you don't start with a hypothesis first its completely meaningless.