|
|
Latent Intercepts-and-Slopes-as-Outco... |
|
Message/Author |
|
|
I am wondering if it is possible to estimate an intercepts-and-slopes-as-outcomes model with latent variables. I have tried the following with a sample of 3,746 level 1 and 126 level 2 units: Usevariables are lf02420, lf02430, lf02470 lf02440, lf02460, lf02500 comm sozibi; between = sozibi; Cluster=idschool; IDVariable is ID; Missing = Blank; Analysis: Type=twolevel random; Processors=4; Estimator = ml; Algorithm = Integration; Integration=montecarlo; model: %within% PFSMw by lf02420*, lf02430, lf02470; PFAMw by lf02440*, lf02460, lf02500; PFSMw@1; PFAMw@1; beta1 | comm on PFSMw; beta1 | comm on PFAMw; %between% PFSMb by lf02420*, lf02430, lf02470; PFAMb by lf02440*, lf02460, lf02500; PFSMb@1; PFAMb@1; comm on PFSMb; comm on PFAMb; beta1 comm on sozibi; comm with beta1; Unfortunately I always receive error messages like this: "THE ESTIMATED BETWEEN COVARIANCE MATRIX COULD NOT BE INVERTED." I have no clue what the problem may be. Did I misspecify something? Or is it impossible to estimate such a model with latent variables in general? |
|
|
Look for close to zero between-level variances. Perhaps some slopes are not random. |
|
Back to top |
|
|