search

stephenslab / susieR

R package for "sum of single effects" regression.
https://stephenslab.github.io/susieR
Other
176 stars 45 forks source link

concern that susie_init_coef does not play nicely with scaling issues #64

Open stephens999 opened 5 years ago

stephens999 commented 5 years ago

If I run a regression y on X1..XL to estimate beta1...betaL, and then supply them to susie_init_coef then I'm worried it will not work well because the beta values are for unscaled Xs. But internally we want the values to be for scaled Xs (assuming susie is invoked with scaling).

We may need to come up with more sophisticated thinking around this.

stephens999 commented 5 years ago

my current thought is that it would be nice to allow user to provide a (sparse) beta_init to susie and handle the initialization internally.