Question re glmfit_multilevel

[Necka, Liz (NIH/NCCIH) [F]]

Hi there,

 

I posted this question on the canlab github, but it might be better suited for this list-serv.

 

I’m trying to verify that the results I get from glmfit_multilevel and that I get from lme4 and nlme in R are the same, but I am finding that they are not. My model has one continuous predictor and one categorical predictor, both on level 1, and I can include an interaction as well.   The main difference seems to be with the categorical variable…  in R, the results for my categorical variable are: B = .49, SE = .41, t = 1.20, p  = .24, whereas in glmfit_multilevel, the results for the categorical variable are: B = .40, SE = .42, t = .97, p = .34… in other words, because the B in glmfit_multilevel is smaller than that in R, and the SE is the same or larger, the T (and therefore p) is very different. For the continuous variable, differences in Beta and SE are probably just due to rounding, though Ts are still a bit different (R: B = 3.79, SE = .19, t = 20.45, p < .001; glmfit_multilevel: B = 3.78, SE = .20, t = 18.70, p < .001).  I’m new to using glmfit_multilevel, but have been using R for years, so am wondering if anyone who has familiarity with both packages can explain how glmfit_multilevel computes betas, SE, and ts that is different from the R packages (I’ve been specifically focusing on lme4).  I’ve included a reproducible example.

 

Thanks in advance.

 

Best,

 

Liz

 

 

 

_______________________________

Elizabeth Necka, PhD

Postdoctoral IRTA Fellow, ANP Lab

National Center for Complementary and Integrative Health

National Institutes of Health

10 Center Drive, 4-1730

Bethesda, MD 20892

(301) 827-1476

Glmfit_multilevelvsR.zip