|
|
 |
| Missing data regression question |
 
|
| Message/Author |
|
|
|
|
I have some missing data in my SPSS data set, and I want to run the model using FIML. I decided to run the Model in Mplus. The issue is that I want a total scale score to do this, so in order to replicate this in Mplus, I made a factor and set all the individual item loadings to one and the regressed the DV on my factor. In SPSS, I accounted for 15% of the variance using total scores, and in Mplus, I accounted for 75% of the variance. I had at least 85% of my observations on each variable, so this radical change seems unlikely. Is there something wrong with my logic or my syntax? Here's the syntax:
MODEL:
factor1 by PEFit9R@1;
factor1 by PEFit15R@1;
factor1 by PEFit24@1;
OHSAT on factor1;
Thanks,
Chris |
|
|
|
|
You might want to use formative indicators; see the Topic 1 handout (towards the end). Fix the coeff's all at 1.
What you do allows measurement error which is not allowed for when summing them up. |
|
|
| Back to top |
|
|
