I know about glmmLasso but would prefer to use elastic net. I wonder if there are any glmm analogues of glmnet out there, or if anyone is working on any. I couldn't find any in my searches. Thanks
$\begingroup$
$\endgroup$
asked Jul 3, 2014 at 17:15
Add a comment
|
2 Answers
$\begingroup$
$\endgroup$
4
Maybe look at the MMS package. The lassop function seems to only be for lasso analysis, yet is clearly built on the glmnet package and includes some of its parameters such as alpha. If nothing else it is likely a good package to modify for your purposes.
-
$\begingroup$ Welcome to the site, @Danny! As you mentioned and is described in the link, "
alpha: argument of ’glmnet’. The elasticnet mixing parameter, with 0 ≤ α ≤ 1. alpha=1 is the lasso penalty, and alpha=0 the ridge penalty." Though the function has the namelassop, as the examples usealpha=0.1, it works as elastic net. In terms of lasso, if I understand it correctly, the advantage of this package v.s.glmmlassoorlmmlassois that the users do not need to select the tuning parameterlambdaby themselves? $\endgroup$– RandelCommented Oct 29, 2015 at 17:13 -
2$\begingroup$ The answer is no. The tuning/regularization parameter is also needed in
lassop(). According to the manual, "mu: Positive regularization number to be used for the Lasso." $\endgroup$– RandelCommented Oct 29, 2015 at 19:34 -
$\begingroup$ Thank you Randel, that's actually quite helpful to me. The missing lambda parameter was what made me unsure. $\endgroup$– DannyCommented Oct 29, 2015 at 19:35
-
$\begingroup$ Yes, the author used a different name for this argument. Interestingly, the argument
alphashows it can do elastic net, but the function name and the descriptions are only related to lasso. Or the package is extended to elastic net but the description is not updated accordingly? $\endgroup$– RandelCommented Oct 29, 2015 at 19:42
$\begingroup$
$\endgroup$
It has been a while, but if anyone is still searching: https://github.com/debbiemarkslab/GELMMnet
