[R-sig-ME] Problems with model (assumptions)

Philipp Singer killver at gmail.com
Tue Nov 24 19:55:14 CET 2015



On 11/24/2015 07:42 PM, Ben Bolker wrote:
> In looking at your model again, I wonder: why are you treating length 
> as a random effect? I would think that lengths would be generally 
> better modeled as continuous covariates, e.g. via an additive 
> model/spline term in the linear model.

Actually, the length is basically the max index of the session, so 
session length.

I thought about whether to include it as a covariate or as a random effect.

I ended up with latter due to the fact that this should help with the 
stabilization of parameters which is specifically true in case of 
imbalance of lengths. Also, when comparing models, the one with the 
random effect performs better.

Maybe I should just incorporate it as a covariate as suggested, the 
inference on the index stays the same. I have actually also done that 
from time to time to speed up things and when crossed random effects 
made troubles.

Thanks!



More information about the R-sig-mixed-models mailing list