I consider a first model where the 6 observables (concentrations of metabolites) are fitted on the data set (the experimental measure of these 6 concentrations). I also have a second model, that is a subset of the first one and models only 4 observables. This second model is fitted on only a part (4 out of 6 concentrations) of the data set used for the first model.

I cannot use the AIC to compare the two models because they are not fitted on exactly the same data. Is these an extension of the AIC or another information criterion that could help me in my case of "nested" models and "nested" data? (If you have a reference, would be great!)

  • $\begingroup$ I don't understand why do you need to compare the models; since they are set on different populations they cannot possibly be compatible with each other. You may want to modify the smaller model with a dummy variable so they both can be set on the same, bigger population. And then you can use AIC. $\endgroup$ – Adam Ryczkowski Apr 25 '14 at 18:02

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.