MLwiN has been able to do this for a long time as it is programmed recursively ; this is what a 1996 (sic) review said

" MLn can analyse up to 15 nested or cross-classified levels which for almost all imaginable purposes is enough"

onlinelibrary.wiley.com/doi/10.1111/j.2044-8317.1996.tb01096.x/pdf

And if you have both programs on your machine you can run MLwiN from 'within' R

http://www.bristol.ac.uk/cmm/software/r2mlwin/

Zhengzheng Zhang, Richard M. A. Parker, Christopher M. J. Charlton, George Leckie, William J. Browne (2016). R2MLwiN: A Package to Run MLwiN from within R. Journal of Statistical Software, 72(10), 1-43.

Hi,

I am not sure how many levels can MCMCglmm handle, but it's one of the common r packages used in continues and categorical variables; you can try it.

Just to say that MLwiN can estimate these models with likelihood, MCMC and iterative bootstrap procedures - and it is comparatively very fast as software reviews have shown.

As always, Thanks Kelvyn Jones.

Thank you Nigus Gebremedhin Asefa . I'll check out the MCMCglmm

Dear Nedim,

xxM, an R package developed by Paras D. Mehta, can handle unlimited levels and more complex cross-classifications. Please check https://xxm.times.uh.edu/ for more detailed information. Hope that helps.

Best,

Yusuf

Thanks Yusuf Kara