It’s another installment in Data Q&A: Answering the real questions with Python. Previous installments are available from the Data Q&A la

Regrets and Regression

submited by

Style Pass

2024-07-02 08:00:05

It’s another installment in Data Q&A: Answering the real questions with Python. Previous installments are available from the Data Q&A landing page.

I want to write a research article that has regression analysis in it. I normalized my independent variables and want to include in the article the results of a statistical test showing that all variables are normal. I normalized using the scale function in R and some custom normalization functions I found online but whatever I do the new data fails the Shapiro Wilkinson and KS test on some columns? What to do?

One is the idea that the independent variables in a regression model have to follow a normal distribution. This is a common misconception. Ordinary least squares regression assumes that the residuals follow a normal distribution — the dependent and independent variables don’t have to.

Another is the idea that statistical tests are useful for checking whether a variable follows a normal distribution. As I’ve explained in other articles, these tests don’t do anything useful.