No. When a case has nothing to contirubute, it cannot be used with FIML or multiple imputation.
Andrea S posted on Wednesday, November 20, 2013 - 10:36 pm
Hi Dr Muthen(s) Im conducting linear growth modeling of my DV (lem) with a time-varying covariate (lse), with data over three time periods (equal time points). Im using version 6 of the software, and my understanding of the defaults of this version is that listwise deletion is off unless specifed, and that the default with missing data should be to use all available data. However, according to the output, I believe the analsyes are only using the data in which participants have data on at least 5/6 measures (lem1-3, and lse1-3), rather than using all available data. This is reducing my sample size significantly. Is there any way to get around this/am I doing anything wrong?
Syntax below: missing are all (-999); USEVARIABLES ARE lem1 lem2 lem3 lse1 lse2 lse3; analysis: type = random; Estimator = ML; MODEL: i s | lem1@0lem2@1lem3@2; lem1 ON lse1; lem2 ON lse2; lem3 ON lse3;
You can change the value 0f the time-varying covariates for those observations that have missing on the dependent variable to any number other than the missing value flag. Then those cases will not be eliminated from the analysis.