Using both raw and weight of evidence values in logistic regression

Question

I am building a logistic regressin model for probability of take-up for a lending product. I have a number of continuous variables. In the past, I have always used EITHER weight-of-evidence transformations OR raw variabels to build such models.

My question is — can one use BOTH WoE and raw values in the same model? I.e. use WoE for variables $x_1,\dots, x_N$ and raw values for $x_{N+1},\dots x_M$?

The model fit looks reasonable, so I am unsure if this is correct.

Can you clarify what you mean by "weight of evidence" variables? Are you asking about transforming only some variables? — gung - Reinstate Monica, Apr 25 '15 at 14:06
thanks for getting back. By "weight of evidence" I mean replacing the raw values of a predictor with the weight-of-evidence value corresponding to it. E.g. if building a model to predict good/bad (credit risk scorecard) using age and income, one can either use the actual age and income values, or the WoE values associated with them. My quesiton is, can one use a mixture of both? — , Apr 26 '15 at 09:29
@danilo276 If you use the same account as you posted from, you can comment in your own posts. Please see about merging your two accounts. — Glen_b, Apr 26 '15 at 11:53
Does this answer your question? Including both transformed and original data (untransformed) in a multivariable linear regression. — kjetil b halvorsen, Apr 23 '20 at 05:17
@gung: This is probably what is meant by weight-of-evidence https://stats.stackexchange.com/questions/462052/intuition-behind-weight-of-evidence-and-information-value-formula — kjetil b halvorsen, Apr 24 '20 at 13:16

score 0 · Answer 1 · answered Apr 23 '20 at 05:16

Weight of evidence (WoE) is really just a specific data transformation, so the same question can be asked for other data transformations. And it has been asked & answered already: Including both transformed and original data (untransformed) in a multivariable linear regression.

Using both raw and weight of evidence values in logistic regression

1 Answers1