[pystatsmodels] Differences between Statsmodels MixedLM and nlme package in R?

[Kevin Nguyen]

Hello all,

I am trying to replicate a collaborator's work, which involved fitting
linear mixed effects models with the nlme package in R. I have set up the
same analysis in Statsmodels with the MixedLM class, and we've confirmed
that I am using the same data, formula, variable coding, and random effects
design. However, my results are persistently different from theirs, with
some inconsistent p-values and coefficients sometimes off by an order of
magnitude.

I have no experience with R myself, so I am unaware if there are any
differences in implementation between R and Statsmodels that would account
for this inconsistency in results. Has anyone with experience with both
packages encountered this discrepancy?

Thanks,
Kevin