The lasso method for variable selection

Question

Whats the benefit of transforming the explanatory variables to have mean 0 and standard deviation 1 in the lasso method, if we didn't do the standardisation what would change?

The answer to https://stats.stackexchange.com/questions/19523/need-for-centering-and-standardizing-data-in-regression/19574#19574 should help. — jbowman, May 03 '18 at 22:40

score 3 · Accepted Answer · answered May 03 '18 at 23:24

If you have multiple features and do regularization(lasso) without normalization, they will be penalized in amounts proportional to their values. You would not want that, and would want them to be penalized on an equal scale relative to each other. Hence you should do that

The lasso method for variable selection

1 Answers1