julia_eval scrambling parameter order when returning :glmerMod

[coverton-usgs]

When returning a GeneralizedLinearMixedModel result using julia_eval I get estimates in the subsequently saved .rds that are out of order with the labels in julia.

The data I am running is quite large and not easily shared. So I apologies for an incomplete example. The code:

julia_assign("m0form", formula(case ~ 0 + Analysisclass + (1|cropyear/individual_local_identifier)))
julia_command(sprintf('Jmodel = fit(GeneralizedLinearMixedModel, m0form, data, Bernoulli(), wts=float.(data.weights), contrasts= Dict(:Analysisclass => DummyCoding(base="aRice_Wet_day")))'))

produces:

But when I save it into R using:

juliamodel<- julia_eval("robject(:glmerMod, Tuple([Jmodel, data]));",need_return ="R") 

I get:

That is a pretty serious discrepancy! Any recommendation on how to prevent this or identify the correct parameter order from the rds alone?

I ran this code in a loop across different datasets and saving only the .rds and not retaining a log to see the original Julia output.

[coverton-usgs]

I also just noticed that the SE estimates are close but not identical between the objects as well