topic Re: surveylogistic fails to converge in Statistical Procedures

surveylogistic fails to converge

Miracle — Wed, 17 Jun 2009 11:42:31 GMT

Hi all.

What should I do in order to get finite odds ratios? I'm using
surveylogistic in order to take into account the clustering effect but
it failed to converge due to quasi-separation.

Please give me some guidance.

Thanks heaps in advance.

Re: surveylogistic fails to converge

statsplank — Wed, 17 Jun 2009 16:25:24 GMT

Hi wong,

I suggest identifying predictor/predictors that cause quasi-separation and removing them from the model.

Regards,
Olga

Re: surveylogistic fails to converge

Miracle — Wed, 24 Jun 2009 03:44:22 GMT

hi Olga.

I'd prefer not do that. Is there any other better approach to address this?
As I know, SAS9.2 proc logistic now has penalized maximum likelihood which can overcome this problem but it cannot be applied in my case since I have to take into account the clustering effect of the individuals from the same household.

cheers