My SEM model includes: f1 by y1-y4 f2 by y5-y8 f3 by y9-y11 f4 by u1-u6 (3-ordered categories each) f5 by f1-f4 1 outcome (ordered categorical) 1 covariate (binary) I use public data and apply panel weights.
1. Running the 2nd order CFA, MI suggest correlating y9 with y11 and y10 with y11. In the modified model these are not estimated (999.000) and the residual variance of y1 becomes negative. What does this mean?
2. MIMIC on f1 with y1 y2 y3 y4 on x@0 (to obtain MI for the observed variables on the covariate), lead to the error message: THE MODEL ESTIMATION TERMINATED NORMALLY WARNING: THE VARIANCE CONTRIBUTION FROM A STRATUM WITH A SINGLE CLUSTER (PSU) IS BASED ON THE DIFFERENCE BETWEEN THE SINGLE CLUSTER VALUE AND THE OVERALL CLUSTER MEAN…… The same occurred when running MIMIC on f2 and f3 (specifying observed on covariate@0). For f4 (u1-u6), once I add the regression of f4 on the covariate (before fixing the paths of observed on covariate), the standardized results table contains just the stdYX and std estimates columns. What might be the reason for these issues?
3. When modeling SEM (2nd order CFA), MI suggest regressing the outcome on f4. Is this a legitimate path?
1. Perhaps you are looking at the standardized results which won't be available due to the negative variance. A negative variance may be an indication of model misspecification. You may want to try bi-factor EFA.
2. The warning is standard with strata that have only one cluster. Why those columns are empty I can't tell without seeing it. Send input, output, data and license number to support.
3. It depends on exactly what your full model looks like. Send input, output, data and license number to support.