The Problem of Correlation: Why you don’t need to have a variable to use it
The Problem of Correlation: Why you don’t need to have a variable to use it
By Gary Angel
|December 18, 2019
The Measurement Minute by Gary Angel
If a model didn’t use gender as a variable, it can’t be gender-biased…right? Totally wrong. Because correlations to protected variables are everywhere in our world, it’s not only easy for an analyst to duck around restrictions on using protected variables, it’s actually pretty hard to avoid doing it.