When should we discretize/binning continuous variables and when not?
My attempt to answer
- In general, we should not bin, because binning will lose information.
- Binning is actually increasing the degree of freedom of the model, so, it is possible to cause over-fitting after binning. If we have a "high bias" model, binning may not be bad, but if we have a "high variance" model, we should avoid binning.
- It depends on what model we are using. If it is a linear mode, and data has a lot of "outliers" binning probability is better. If we have a tree model, then, outlier and binning will make too much difference.
Am I right? and what else?
- I thought this question should be asked many times but I cannot find it in CV only this one: Should we bin continuous variables?. The answer is really short. I wish we have a long answer or formal references.
- Feel free to mark this question as too general or duplicate. I am still learning how to ask..