> especially when n=15(if this was a statistical study on insulin response to artificial sweeteners, people would be saying the sample size is too small to draw conclusions)
And those people would be wrong. It's incorrect to dismiss a study based on sample size without a discussion of significance and effect size in the context of the data.
Moreover I reject the premise that you should be assessing this story quantitatively rather than qualitatively. But if you insist: what are your priors on whether or not a given company engaged in discrimination, and how do these change if you're told 75% of employees of a particular demographic stated there was discrimination?
Okay - one context is where one department has a large number of black people, and that department is defragged for purely business reasons to a different office across the country. Most people never choose the relocation package, and so a big chunk of black employees end up quitting. That's the event that triggered 8 black employees of the 15 who left as I understand it from the article.
> what are your priors on whether or not a given company engaged in discrimination, and how do these change if you're told 75% of employees of a particular demographic stated there was discrimination?
I completely agree that it's a very bad look. It is probably even more likely than not that given those facts, it is due to racism. I guess the question comes down to, philosophically, how one answers the following question: In the quest to eradicate racism and racists, is it better to be over-zealous and destroy a few non-racists to make sure you get all the actual racists(the chemo approach) - or is it better to be slightly more circumspect and let a few racists slip through the cracks so that far fewer non-racists are punished (the US judicial system ideal)?
But this question assumes that the root cause of the problem is that Coinbase has a disproportionately high number of "bad apples." To me that's exactly backwards; it seems much more likely that institutional culture tolerates and even encourages racism.
Yes that's what was asked; which ones?
> especially when n=15(if this was a statistical study on insulin response to artificial sweeteners, people would be saying the sample size is too small to draw conclusions)
And those people would be wrong. It's incorrect to dismiss a study based on sample size without a discussion of significance and effect size in the context of the data.
Moreover I reject the premise that you should be assessing this story quantitatively rather than qualitatively. But if you insist: what are your priors on whether or not a given company engaged in discrimination, and how do these change if you're told 75% of employees of a particular demographic stated there was discrimination?