Too many iterations/collinearity during poisson regression
Scott Raynaud
Trivellore Raghunathan
Are these binary variables or count variables? If a binary variable is declared as count then it can create imputed values any integers and may result in the convergence problems.
Raghu
I have about 50 variables on which I'm performing imputation. If I treat these as logistic regressions everything goes fine, but declaring them as count variables always throws an error: too many iterations/collinearity during poisson regression. I've tried restricting to the single most important predictor and increasing iterations to no avail. Any other suggestions?
--
You received this message because you are subscribed to the Google Groups "IVEware" group.
To unsubscribe from this group and stop receiving emails from it, send an email to iveware+u...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
Scott Raynaud
Ok, I get it. They need to be frequencies. My data are from a repeated cross sectional survey. My substantive model predicts health outcomes controlling for sample design information (PSU, stratum, sample weight), year of survey and other demographic variables like region and education. Should I segregate my frequencies by PSU, stratum, year and demographics? This could lead to small cells.
And what if I choose to run my substantive model as a negative binomial? Do you think I face significant questions about congeniality between my imputation model and my substantive model?
Trivellore Raghunathan
You need to expand your frequencies or grouped data into individual level records with dummy variables and then use iveware.
You can then group the imputed data and fit whatever model.
Raghu