|
|
 |
| Accounting for Missing Data in Longit... |
 
|
| Message/Author |
|
|
|
|
I am trying to run a longitudinal SEM with latent constructs. I have a lot of missing data (around 3-7,000 cases of missing) with a large sample size (around 20,000). The error is: WARNING: THE LATENT VARIABLE COVARIANCE MATRIX (PSI) IS NOT POSITIVE DEFINITE...
I have tried using an ML, coverage=0, trying other robust estimators,etc.
Do I need to try something other than an ML estimator to account for the missing data in this type of structure?
MODEL 1:
DATA: ...
VARIABLE:
MISSING ARE ALL(999);
ANALYSIS:
COVERAGE=0;
ESTIMATOR=ML;
MODEL:
cog10 BY totre10 ser710 verbf10;
cog12 BY totre12 ser712 verbf12;
cog14 BY totre14 ser714 verbf14;
cog16 BY totre16 ser716 verbf16;
i s| cog10@0 cog12@1 cog14@2 cog16@3;
OUTPUT:
SAMPSTAT STDYX;
MODEL 2:
DATA:...
VARIABLE:
MISSING ARE ALL(999);
ANALYSIS:
COVERAGE=0;
ESTIMATOR=ML;
MODEL:
cog10 BY totre10 ser710 verbf10;
cog12 BY totre12 ser712 verbf12;
cog14 BY totre14 ser714 verbf14;
cog16 BY totre16 ser716 verbf16;
cog12 ON cog10;
cog14 ON cog12;
cog16 ON cog14;
OUTPUT:
SAMPSTAT STDYX; |
|
|
|
|
| We need to see your full output - send to Support along with your license number. |
|
|
| Back to top |
|
|
