[R] clmm model specification

Oscar Smallenbroek osmallenbroek at gmail.com
Wed Apr 16 19:21:40 CEST 2014

Hello All,

I'm trying to fit a cumulative link mixed model clmm(). I'm currently
having issues with the diagnosing what is wrong with my model from the
output I am getting. I've attached the output in PDF.

The model:
I'm estimating the effect of education, social status, class of parents and
control variables on the educational attainment of respondents (eduRes).
Education and class are ordinal variables. The mixed effects are country
level and describe the education system.

The problems:
I've identified a few sources of error. First I am using an SPSS data set
which gives me some errors when I load it in (see pdf). Second I could be
miss-using the commands in R because I am new to it.
Thirdly, the Hessian is above 1e4.
Fourth, the output I got from my null model changes the variable names (no
idea why).
This also happened the first time I tried fitting the model. I fixed this
by chancing the contrast for eduPar, eduRes and the EGP (all the ordinal
variables). These variables now seem to have the right names. However the
cohort and gender dummies do not.
Lastly, when I add my country level variables the hessian does not compute.

Any suggestions are welcome.


-------------- next part --------------
A non-text attachment was scrubbed...
Name: Rquestions.pdf
Type: application/pdf
Size: 83356 bytes
Desc: not available
URL: <https://stat.ethz.ch/pipermail/r-help/attachments/20140416/9c88c8bf/attachment-0002.pdf>

More information about the R-help mailing list