0

I would like to estimate a mixed logit model for a group of 80,000 individuals and 1,200 alternatives in R. Packages like gmnl and mlogit take a significant time to estimate the model, even without including random parameters. The mnlogit package allows the estimation of models more efficiently and quickly. However, it does not allow to include random parameters. That is why I would like to know if there are packages (with available examples) designed for estimating Mixed Logit models with large datasets. So far I have found the mixl package. However, I have not found any examples available to help me understand how the package works.

martin sielfeld
  • 1
  • 1
  • 1

1 Answers1

0

Assuming you want frequentist methods, the usual suspects in R would be glmmTMB or lme4.

If you're willing to try Julia, then MixedModels.jl can handle logistic models that size. With JellyMe4.jl, you can even move the Julia fit back to R to use the R package ecosystem for examining the model (packages like car, sjtools, etc.).

Livius
  • 3,240
  • 1
  • 17
  • 28