I am using a two-stage approach (whereby admissions to hospital are nested within patients) to account for multiple admissions of individual patients during a study period. Intuitively it makes sense that my effect estimate would be lower in this case as individuals can have multiple admissions without developing the outcome of interest. However I can't quite explain why this is the case. Any help would be greatly appreciated!
I guess you are using a mixed effect model, right? If yes, this is a well-known feature of generalized linear mixed effect model (GLMM) vs marginal models (like "regular" logistic regression ). See any good text book on GLMM, eg Diggle et al 2002 Longitudinal data analysis, OUP.
Sorry it's a multi-level model rather than mixed effects. So admissions are nested (level 1) within patients (level 2). I haven't found any literature that gives me a clear explanation as to why my effect estimate would be lower compared to using GLMs for Binary outcomes. Cheers!
!i think it's a special case of a mixed-effect model (and consequently the reason why you observe this phenomenon). With which software do you fit it ? Can you write it in a symbolic way ? I will elaborate further if you confirm my guess.
Yes it is a multilevel mixed-effects logistic regression model. I used stata xtmelogit command.
- Badges
- Science topic