Last 3 Days
Multiple Imputation H1 - estimator ch...
Missing Data Modeling
posted on Friday, August 21, 2015 - 10:10 am
Dear Mplus Team,
I note that if no estimator is specified for an H1 imputation then for the following model...
CATEGORICAL: v2 v3;
AUXILIARY = v4;
IMPUTE= v2 (c) v3 (c) v4;
NDATASETS = 10;
SAVE = LONG_2015_*.dat;
TYPE = BASIC TWOLEVEL;
...this is produced in the ouput as default:
Information matrix OBSERVED
...Obviously if I choose ESTIMATOR = WLSMV then this output changes to reflect WLSMV but what exactly is this estimator influencing as I understood that H1 imputation uses BAYES.
Is BAYES used for the imputation part of the model regardless of ESTIMATOR specified with the choice of ESTIMATOR being used to generate sample statistics? I'm confused here.
1) Which ESTIMATOR choice should I specify for H1 imputation (if any), keeping in mind that my follow up analysis will use WLSMV since my data is categorical.
Bengt O. Muthen
posted on Friday, August 21, 2015 - 2:45 pm
With Type=Basic the estimator only influences the sample statistics, not the imputation. The estimation used for the imputations is Bayes. This is why UG ex 11.5 doesn't mention an estimator.
See also the diagram on page 516 of the UG.
Back to top
Add Your Message Here
This is a private posting area. Only registered users and moderators may post messages here.
Enable HTML code in message
Automatically activate URLs in message