0

I've seen this statement several times, but I haven't seen an explanation for it. What about Lasso Regression makes it struggle against multicollinearity?

When I've seen this stated, it's usually in the context vs. Ridge regression or some feature selection method like subset/forward/backward selection, etc...

student010101
  • 334
  • 2
  • 10
  • 1
    It's hard to tell, without context, what someone might mean by "struggle with" or "struggle against" multicollinearity. Could you provide a source to help us understand? – whuber May 06 '21 at 20:43
  • 1
    @whuber Here is one https://stats.stackexchange.com/questions/4272/when-to-use-regularization-methods-for-regression. The answer by chl states "To achieve sparsity, the lasso is more appropriate but it will not necessarily yield good results in presence of high collinearity (it has been observed that if predictors are highly correlated, the prediction performance of the lasso is dominated by ridge regression)" but I don't know what this is referring to. I've seen this in some other posts as well, but I don't think I've seen any context surrounding it. – student010101 May 06 '21 at 21:10
  • The impression that I've gotten is that this is something that's observed in practice, and there doesn't seem to be a theoretical reason for it. – student010101 May 06 '21 at 21:10

0 Answers0