Dear RG Members,
I am testing the factor structure of some psychological scales. The participants have some missing values in their items answers.
Is it possible to use the multiple imputation procedures of Stata in combination with exploratory (EFA) and confirmatory factor analyses (CFA with SEM)?
Thank you.
Andrés.
Not that I am aware of. help mi_estimation##estimation_command will give you the estimation approaches available for mi
To some degree, imputing missing data while testing psychometric properties of an instrument doesn't make much sense. After all, these procedures are looking for some patterns in the actual responses, and if there are missing values that need to be imputed (with some degree of error), then the psychometric properties will be based on inaccurate data. Moreover, we haven't even touched upon the important point of why are the data missing and if they are missing completely at random, missing at random, or biased.
Thank Dr. Linden. Indeed, we are planning to do factor analysis with participants with full answers; but as a sensitivity analysis, we would like to use also multiple imputation procedures. Best regards.
Got it. Perhaps you could consider running simulations: (1) generate missing data, (2) multiply impute, (3) run factor analysis, (4) test sensitivity?
I
- Badges
- Science topic
- Similar topics
- Psychology