Closed
Description
I find the documentation on f-regression feature selection very obscure, and would suggest the following:
- make it clear that the f-regression feature selection is just a univariate linear model, where features are ranked accordingly to the significance of the regression parameter. This is actually equivalent to ranking features according to their correlation with the target variable.
- make it clear that this is not an iterative procedure to include features (ie not what is described here: http://stats.stackexchange.com/questions/204141/difference-between-selecting-features-based-on-f-regression-and-based-on-r2).
- maybe mentionned that this is akin to sure independence screening? http://orfe.princeton.edu/~jqfan/papers/06/SIS.pdf
Note that all of this may be trivial for anyone more familiar with the feature selection literature.