search

pbreheny / grpreg

Regularization paths for regression models with grouped covariates
http://pbreheny.github.io/grpreg/
34 stars 14 forks source link

Scaling in standardize.c #21

Closed morris16206 closed 5 years ago

morris16206 commented 5 years ago

I found that when doing standardization for the design matrix, the code is using degrees of freedom = n for standard deviation. I am wondering why not use n-1, which is commonly used in other standardize function and should give an unbiased estimator for the variance.

pbreheny commented 5 years ago

Standardization based on root-mean-square is universally used in machine learning. Whether or not there is a bias for the variance is irrelevant, since no variances are being estimated here.