4

When does LASSO fail in terms of:

  1. out of sample forecasting?
  2. selecting relevant variables?

(If there is a distinction between 1 and 2 to begin with.)

Also,

  1. Under what conditions would one be better off using ridge regression or other techniques instead of LASSO?

I am looking for the most salient examples and conditions which should ring a bell for an applied modeller.

Richard Hardy
  • 54,375
  • 10
  • 95
  • 219
  • From the top of my head, I remember reading about LASSO failing in cases when covariates are highly correlated. Source: Trevor Hastie et al., "Statistical Learning with Sparsity" https://www.crcpress.com/Statistical-Learning-with-Sparsity-The-Lasso-and-Generalizations/Hastie-Tibshirani-Wainwright/p/book/9781498712163 – Vladislavs Dovgalecs May 23 '16 at 15:43
  • @xeon, that is what the answer by Scortchi says in the linked thread. Good that you agree. – Richard Hardy May 23 '16 at 15:50
  • I accidentally found more similar threads. One http://stats.stackexchange.com/questions/140544 Two http://stats.stackexchange.com/questions/7935 Could you review them and check if they perhaps should be closed as duplicates (or maybe merged), or edited to make the titles more distinct? – amoeba May 24 '16 at 13:39

0 Answers0