LASSO - NegBin/Quasi Poisson in Matlab

#1
Hello,

I am currently attempting to implement variable selection using LASSO. Unfortunately, since my data is frequency and generally overdispersed, I would need to use either Negative Binomial or Poisson regression. I know there is a function in MATLAB (lassoglm) which would allow L1 regularization of most applicable distributions (but not neg binomial or quasi-poisson). Is there any way I could implement LASSO based on this data?

Any help is appreciated.
 

Lazar

Phineas Packard
#2
Does glmnet for matlab allow for quasipoisson? if not the other option is Bayes approach to lasso as you can add a parameter for overdispersion.
 
#3
I think my approach will be to use the in-house Matlab function lassoglm for poisson regression. Then, I'll do an adjustment to the p-values, standard errors by doing dispersion*vcov matrix.