Sexual fetishes on the political compass, men and women. Total sample size was over 19,000!
679
3,895
14.8K
Conversation
Replying to
Aren't most r^2 < 5 %? It can easily be statistical noise.
Or am I missreading it?
1
5
Replying to
For almost all studies like this, you should expect it to be statistical noise, yeah! But the significance depends on sample size; the larger the sample, the more likely it is that your finding isn't just statistical noise.
Replying to
The likelihood ratios (how much more likely it is to get the observed correlation vs the null) in my graphs ranges from 5 to around 1e+10 (extraordinarily likely)
1
19
Show replies
Replying to
Actually, when your dataset is so huge, then it's bound to be statistically significant every time, no matter the findings. You should have drawn a random sample which would be much smaller in order for it not to be bs.
Replying to
The highest correlation is 0,12. Even if you had 5× bigger sample, it wouldn't show any causality despite it's low p value.
Also the bigger male sample has 2× smaller the highest correlation. That implies something.