I've read that univariate analysis is a huge part of EDA, pretty much everywhere. What I understand of univariate analysis basically comes down to using Seaborn pairplot and correlation matrices and heatmaps.

However, today I read about "Simpson's Paradox" on a kaggle kernel, then in a Britannica article. It was quite fascinating and seemed to render useless what I believed to be commonly accepted best practices.

I then read that one way to combat this sort of misleading analysis is through partial residual plots:

$$\text{Residuals} + \beta_iX_i \text{ vs } X_i$$

What I can't figure out is how this formula controls for multiple independent variables in an intuitive way. What's the explanation for how this works?



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.