www.cmm.bristol.ac.uk/forum

Dear all,

I want to conduct a two-level nested model using longitudinal data (level1 = observation, level2=ID).
My dependent variable is a binary outcome.

When I try to fit my nullmodel using 2nd order PQL (based on the starting values of 1st order MQL), my model does not converge.
I am not sure why...

Any help with this would be much appreciated!

Katrijn

2nd Order PQL will use the higher-level residuals when fitting the model, so it's probably worth taking a look at them to see if anything looks odd. It would also be worth fitting the model with MCMC to see whether this has any problems.

Agreed with ChrisCharlton's suggestion of running MCMC. I also encountered similar issue for a four-level random intercept model with binary outcome. MCMC worked well while the 2nd order PQL had an issue of convergence.
Thanks,
YongJoo