Covariance structure in GAMLj Mixed Model?

General help and assistance with jamovi. Bug reports can be made at our issues page: https://github.com/jamovi/jamovi/issues . (If you're unsure feel free to discuss it here)

by Asia » Tue Aug 11, 2020 12:03 pm

I have a question about covariance structure used by default in GAMLj Mixed Model. On the website: https://gamlj.github.io/mixed.html I found information that if we want to obtain a variance component (VC) model we should select "Not correlated" option in "Effects correlations" . Do I undestand correctly that by default this modul assumes unstructured covariance structure (UN) ?

Thanks a lot for all your help!
Asia
Asia
 
Posts: 1
Joined: Tue Aug 11, 2020 11:50 am

by mcfanda@gmail.com » Mon Aug 17, 2020 11:25 am

Yes, you assume well. GAMLj uses the same default that the underlying R lmer() command. Thus, UNstructured by default, meaning all random coefficients are allowed to be correlated.
User avatar
mcfanda@gmail.com
 
Posts: 260
Joined: Thu Mar 23, 2017 9:24 pm

by ganzz_gut » Mon May 17, 2021 9:32 am

I am a relative newbie to Mixed Models, with this perhaps naive question... If I use the Random Intercept Model then setting covariance structure (i.e. correlated or uncorrelated) does not matter, doesn't it? As the covariance is assumed for random components (i.e., more than one). While in the case of Random Intercept Model there is only one component, which is cluster variable?
ganzz_gut
 
Posts: 1
Joined: Mon May 17, 2021 9:17 am

by mcfanda@gmail.com » Tue May 18, 2021 7:01 am

Correct.
however, the one component is not the cluster variable but it is the intercept.
User avatar
mcfanda@gmail.com
 
Posts: 260
Joined: Thu Mar 23, 2017 9:24 pm


Return to Help