For multiple linear regression problem, I have both categorical and numerical variables in the data. I am checking the correlation for numerical variables for EDA and standardizing them by taking log. For rest of the categorical variable columns contains 2 values (either 0 or 1). So I wanted to understand if we should consider categorical variables in the correlation matrix alongside numerical vars and keep them as is with log-transformed numerical variables in input for the regression model. Please guide.
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
https://stats.stackexchange.com/questions/484299/how-to-check-the-correlation-between-categorical-and-numeric-independent-variabl/484300#484300 This link mentions a couple of approaches (albiet in R, need to look for corresponding implementations in Python). These approaches assume the categorical variable to be dichotomous, which is true in your case
1.https://en.wikipedia.org/wiki/Point-biserial_correlation_coefficient Point biserial correlation
For Python- please check this https://www.statology.org/point-biserial-correlation-python/
2.http://www.sthda.com/english/wiki/unpaired-two-samples-wilcoxon-test-in-r
Please let me know, it this was helpful
