student07 posted on Saturday, July 28, 2007 - 6:11 am
I have a question of understanding regarding the "type = twolevel complex" command as described on p. 401. I wonder whether this command can be used to include covariates which refer to the primary sampling unit.
Example: Assume a data structure with employees (level 1, N= 3000) sampled from teams (level 2, N=60) sampled from companies (level 3, N=30).
Objective: I intend to model a two-level SEM with covariates on the within-teams (level 1)and between-teams level (level 2). This works fine, results are OK.
However, I also have an interesting covariate "X3" which has been measured for companies only - thus, conceptually a level-3 covariate.
Question: Would it be feasible to regress some between-team latent variable (level 2) on "X3" and obtain correct estimates for "X3" by using the
"Type = twolevel complex" "Cluster is companies teams" ?
Jane Wu posted on Thursday, January 03, 2013 - 9:08 pm
A related question: If specifying TYPE = COMPLEX to run a single-level SEM (taking into account the nesting of employees within teams), can you confirm that I can't test team-level (level 2) effects on employee (level 1) outcomes?