I am having problem deciding wheter to define a variable as a random effect to include in our logistic regression model. Any help on this subject would be most appreciated.

In this model, we are trying to determine the effect of region of residence on different endpoints after hospital admissions. We have included several demographic and hospital variables as fixed effects as well as region.

Due to the fact that most people from regions are hospitalized in their regional hospital, each hospital is highly correlated with region.

Is it appropriate to use hospital (as a categorical variable) as a random effect to control for the potential bias in the varying quality of care between hospitals?

  • $\begingroup$ first impression for me is that it should use hospital as a random effect, which would allow you to draw conclusions that are generalised across hospitals, but I may be wrong $\endgroup$ – rg255 Jun 1 '16 at 11:23
  • $\begingroup$ First of all you need a reading of logistic regression. $\endgroup$ – Subhash C. Davar Jun 1 '16 at 12:13
  • $\begingroup$ If hospital & region are nearly the same variable, then saying anything much about both or separating the two effects (whether as fixed or random effects) will be difficult. Perhaps it is not really necessary to have both in the model and you could just use the one that is a-priori considered to be logically more influential? $\endgroup$ – Björn Jun 1 '16 at 13:41
  • $\begingroup$ You are interested in effect of region, not effect of hospital --> you only want coefficients per region for comparisons. So in all cases you certainly dont want to have hospital as fixed effect. Maybe you should test model with it as random, model without it and carefully assess fitting statistics. Note that I can imagine there is clearly way to assign to hospitals additional characteristics such as size or location near/far from big cities that may interest you more in your investigations $\endgroup$ – Eric Lecoutre Jun 1 '16 at 14:15

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.