|
|
Multivariate multilevel censored model |
|
Message/Author |
|
R McDowell posted on Monday, August 22, 2016 - 8:48 am
|
|
|
Hello. I am trying to fit a multivariate (2 outcomes) multilevel model where both outcomes are treated as censored normal (below). If I fit the model without censoring, a correlation between the two outcomes at the within level is reported as default, but not when the censoring is added. If I explicitly add %within% a with b; I get an error. I am able to specify a within-cluster correlation if I use a latent variable formation instead. %within% F by a b; Is it sensible/practical to try and incorporate such correlations in a multilevel model where both outcomes are censored normal? Thank you |
|
|
Yes, the factor approach is the way to go here. |
|
R McDowell posted on Wednesday, August 24, 2016 - 2:42 am
|
|
|
Many thanks for confirming that, it's a great help to know this. |
|
Back to top |
|
|