I am using the PISA 2015 data and trying to run a mixed-effects ridge and lasso regression model . Schools (i.e. school id) will be included as a random effect, and student-level (e.g. motivation, socioeconomic status, epistemic beliefs) and school-level (e.g. percentage of students from disadvantaged families, student-teacher ratio, professional development availability) predictors will be included as fixed effects. The response/outcome variable is scientific literacy scores of students.
Just to give a context to why I'm specifically asking about lasso and ridge regression; I want to compare the predictive accuracy of several methods, and two of the methods happen to be lasso and ridge regression. However, I'm having trouble finding packages that can accommodate such an analysis.
I am using R to run my analyses. I have tried using lmmlasso::lmmlasso to specify the mixed-effects lasso model (it is the only package I found that does such an analysis), but found that there is a problem/error with this package. "This error comes from the package penalized which lmmlasso is built on, which has apparently been classified on CRAN as ORPHANED" (see https://github.com/yonicd/lmmen/issues/3#issuecomment-478418540). I have searched for a package that runs mixed-effects ridge models, but to no avail...
My question is - Does anyone know of (alternative) packages I can use to run mixed-effects models with lasso penalty and ridge penalty in R?
Henry Ratul Halder Hey Henry, thank you for your suggestion. However, this does not really help with my issue of running a mixed model with lasso penalty and ridge penalty specifically.
This might help a bit
https://m-clark.github.io/posts/2019-10-20-big-mixed-models/#r-packages-for-mixed-models-with-large-data
Hi Shermain Puah ,
I am hoping you might have found an adequate package for your purpose by now, if not, I am just sharing two that I came across during my research:
1. https://github.com/jklosa/seagull
This package provides regularization paths for the lasso, group lasso, and sparse-group lasso. The underlying mathematical model is a mixed model.
2. Another one with Lasso penalty: https://cran.r-project.org/web/packages/glmmLasso/glmmLasso.pdf
Both of them, however, doesn’t have ridge penalty, unfortunately! But I hope it will be of some use.
Good luck!
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Statistical Software
- R