Posthoc power analysis
Message/Author
 jeon small posted on Friday, March 01, 2013 - 8:05 pm
How do you perform a posthoc power analysis. I have a sample of 438 girls and boys. We used 11 indicator variable to form four variables. The fit of the structural model was satisfactory with CFI=.954, RMSEA=.05, DF=40 CHI SQUARED=95.85 PROBABILITY=.0000
The standardized regression weights were significant and ranged in size form .65-.03.
 Linda K. Muthen posted on Saturday, March 02, 2013 - 2:30 pm
See Example 12.7 where the values from a real data analysis are saved and used as population parameter values for a Monte Carlo simulation study that can examine power. This method has its critics. You may want to explore this issue on SEMNET.
 jeon small posted on Monday, March 04, 2013 - 1:38 pm
I ran this model in Amos using maximum-likelihood estimation (for the missing data)--however, I can not get the model to converge in Mplus. What am I doing wrong? My output is below.

STRUCTURAL MODEL OF ADOLESCENT ALCOHOL USE AND PERCEPTION OF IPV

SUMMARY OF ANALYSIS

Number of groups 1
Number of observations 553

Number of dependent variables 10
Number of independent variables 0
Number of continuous latent variables 4

Estimator ML
Information matrix OBSERVED
Maximum number of iterations 1000
Convergence criterion 0.500D-04
Maximum number of steepest descent iterations 20
Maximum number of iterations for H1 2000
Convergence criterion for H1 0.100D-03

Input data file(s)
h:/M_russell/mplus10_08_12v1.csv

Input data format FREE
 Linda K. Muthen posted on Monday, March 04, 2013 - 3:37 pm
 Christoph Schaefer posted on Monday, January 15, 2018 - 6:07 am
Dear Professors Muthen,

I would like to estimate the power to detect that a path coefficient is different from zero and am trying to use
https://www.statmodel.com/power.shtml
as a procedure. Alas it hasn't worked so far.
A simplification of my model looks like the following, with three IVs and three DVs. Two of those are latent, four manifest:

IV1 by
Indicator1
Indicator2
Indicator3;

DV1 by
Indicator4
Indicator5
Indicator6;

DV1 on
IV1
IV2
IV3;

DV2 on
IV1
IV2
IV3;

When I follow step 1 of the procedure on
https://www.statmodel.com/power.shtml, fixing values for the factor loadings and the path coefficients, the resulting covariance matrix of the residuals contains zeros, which is not accepted by Mplus in step 2.
Do I have to fix more in step 1 than the factor loadings and the path coefficients?

(In step 1, I have used the most simple matrix possible, in the form of:
0 0 0 0 0 0 0 0
1
0 1
0 0 1
0 0 0 1
0 0 0 0 1
0 0 0 0 0 1
0 0 0 0 0 0 1
0 0 0 0 0 0 0 1
)
 Bengt O. Muthen posted on Monday, January 15, 2018 - 7:57 am