GAMLj - plotting Random Effects in generalized mixed models

Everything related to the development of modules in jamovi

by mcfanda@gmail.com » Mon Feb 10, 2020 9:01 am

Yes, more than statistical the reason is practical. The lmerTest::ranova() function in the R lmerTest package is not defined for GLMM. The plan is to implement the random variances test from scratch for GLMM, but that takes time.
User avatar
mcfanda@gmail.com
 
Posts: 251
Joined: Thu Mar 23, 2017 9:24 pm

by shabde » Wed Mar 17, 2021 4:20 pm

Its me again - again having issues with plotting interactions between factor and covariates in GAMLJ mixed model.

@McFanda maybe this is now the LMER issue you were talking about last year:

Have a look at the attached LMM. The fixed effects estimations are clear and reasonable and according to what would be expected in theory. The slope of the redundant "Trompete" category of the factor "content" should be negative, (since all dummy terms in the regression become zero, we are left with a negative slope of "umhuellung") but, however, is somehow plotted as slightly positive by GAMLJ. The same weird behavior can be seen in the estimated means. This should not be the case, or am I missing something?

Moreover, the model has just a random intercept, but the plot looks as if there were freely varying random slopes.

In this dataset, the factor "content" is not balanced / hierarchical, which might be related to the problem.
Attachments
weird_plot.omv
example lmm
(213.47 KiB) Downloaded 63 times
shabde
 
Posts: 12
Joined: Thu Aug 29, 2019 7:53 pm

by mcfanda@gmail.com » Wed Mar 17, 2021 10:47 pm

No statistical reason. It's more a matter of time and (cognitive) resources. It'll come nonetheless.
thanks for the support :-)
User avatar
mcfanda@gmail.com
 
Posts: 251
Joined: Thu Mar 23, 2017 9:24 pm

by shabde » Thu Mar 18, 2021 8:59 am

@McFanda

Let me know when I can help you somehow in debugging. I think GAMLJ is the most excellent of all the Jamovi modules!
shabde
 
Posts: 12
Joined: Thu Aug 29, 2019 7:53 pm

by mcfanda@gmail.com » Thu Mar 18, 2021 3:41 pm

Hi
thanks, good to know. As for the LRT, there's no statistical reason. It's just one of the many options we are working on. It'll come soon
User avatar
mcfanda@gmail.com
 
Posts: 251
Joined: Thu Mar 23, 2017 9:24 pm

by shabde » Thu Mar 18, 2021 4:28 pm

Hi @McFanda -
Your reply confused me a liittle. I think you accidently did not read my post from Wed Mar 17, 2021 4:20 pm. It was not about LRT at all. There seems to be some weird bug (mabye a new issue) when plotting estimated means. Check the example file that I posted and my explanation.
shabde
 
Posts: 12
Joined: Thu Aug 29, 2019 7:53 pm

Previous

Return to Module development