Esther on Nostr: One thing that studying actual statistics has made very clear is how utterly shit ...
One thing that studying actual statistics has made very clear is how utterly shit much of "data science", especially in the tech industry, is.
I've seen so many instances of people in "data" teams running A/B-Tests (which are mostly non-consensual psych experiments and therefore unethical in the first place) and then just going "well, the average of metric X is a little higher in group B so let's do that" with zero regard to effect sizes, significance, test power, variance analysis, … not even a simple t-test.
I've seen so many instances of people in "data" teams running A/B-Tests (which are mostly non-consensual psych experiments and therefore unethical in the first place) and then just going "well, the average of metric X is a little higher in group B so let's do that" with zero regard to effect sizes, significance, test power, variance analysis, … not even a simple t-test.