[R-sig-ME] MIxed Effects Logistic Regression error "rank deficient" #534

[Phillip Alday]

Rank deficiency here means that not all all combinations of all
(fixed-effects) predictors appear in the data and so the model matrix is
incomplete (not full rank) and not all coefficients / fixed effects can
be estimated. If you have a design where you expect this, this is not a
problem. But if all combinations should be present, then this suggests
that your data are not what you think they are. Maybe double check that
you didn't accidentally change the data somehow or incorrectly code a
categorical variable?

Phillip

On 22/7/19 11:29 pm, Canarte Gutierrez,Climaco D wrote:
> Hello
> 
> I am running mixed effects logistic regression that is throwing the following error: "fixed-effect model matrix is rank deficient so dropping 1 column / coefficient"
> 
> This model is the full random-intercept model with all level-1 and level-2 predictors (20 IVs in total.. yeah is a big one) . I did not have this error when gradually adding level-1 and level-2 predictors.
> 
> After changing one categorically variable I recently created with its former version, the full-random intercept did not show the “rank error”, but the b estimates and p-values changed.
> 
> Does anyone have an idea what's going on? Should I disregard the results of the model with the "rank deficient"  error "?
> 
> Thanks
> 
> David Cañarte
> Ph.D. Candidate
> Department of Sociology
> and Criminology
> University of Florida
> dcanarte using ufl.edu<mailto:dcanarte using ufl.edu>
> 
> 
> 
> 
> 	[[alternative HTML version deleted]]
> 
> _______________________________________________
> R-sig-mixed-models using r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>