Muthen posted on Wednesday, Febru3:02 pmĪs far as I understand, combining estimates and s.e.'s from analyses of multiple-imputed data could be done in the usual way also when using WLSMV. My question is whether the coefficients and standard errors generated by the Mplus WLSMV estimator present particular problems for those planning on combining results from several separate imputed data sets.īengt O. I have data that do not fit the assumptions Mplus imposes for SEM with missing data so I am using a multivariate, multiple imputation approach such as that advocated by Little and Rubin. You can try to sharpen the convergence criterion as described in the User´s Guide.Īnonymous posted on Wednesday, Febru9:07 am

The H1 estimation that this leads to can be difficult if there are large percentages of missing data - see the Covariance Coverage output. You may want to first to a type=basic missing. If you give the Mplus statement type=missing h1, the program first does H1 and then H0. Am I missing some fundamental piece of the on Monday, Ma8:18 am It runs fine when I do not ask for 'type= missing h1 ' but then I can't get the chi-sq. I am having difficulty getting Mplus to converge on H1 (and thus to get a chi-sq test) for a missing-data-latent-growth-curve model, even when I fiddle with the starting values and convergence criteria. Matthew Archibald posted on Saturday, Ma5:06 pm Each parameter is estimated directly without first filling in missing data values for each individual. It uses all data that is available to estimate the model using full information maximum likelihood. No, Mplus does not impute values for those that are missing. Mplus Discussion > Missing data Mplus Homeĭoes Mplus impute values for those that are missing?

Stata 13mp serial number