Normality is very essential for Linear models specially for the independent variables.
However during one of the projects i had one of the stakeholders ask if for binary classifiers or classification models like logistic or forest does normality in independent variable also hold good.
The outcome variable is dichotomous or may have 3 categories in some cases hence the independent variable being non normal may not impact the outcome.
Request experts in stats to give some more details and explanation as to why it could be beneficial or non beneficial to have a non normal variable in the model.
Thank you.
Normality is not an essential condition for independent or dependent variables in Linear Models. The only thing to check is the normality of error distribution, which itself is a weak normality assumption. Logistic regression itself is not strictly a non-linear model however, the error distribution is not required to be normal. As far as independent variables are concerned, they may or may not be drawn from Normal distribution.
- Badges
- Science topic
- Similar topics
- Data Mining
- Classification