I'm trying to perform multiple imputation for a two level longitudinal analysis. I have missing on both DV and covariates. Every subject is measured on 3 variables (SCORE1L1 SCORE2L1 SCORE3L1) at several timepoints (TIME) . I also have three level-2 covariates (AGE SEX SES).
My syntax is the following:
VARIABLE: NAMES = SubjectID TIME AGE SEX SES SCORE1L1 SCORE2L1 SCORE3L1 GROUP1 GROUP2; USEVARIABLES = TIME AGE SEX SES SCORE1L1 SCORE2L2 SCORE3L1; cluster = SUBJECTID; between = AGE SEX SES; within = TIME; auxiliary = GROUP1 GROUP2; Missing = .;
DATA IMPUTATION: IMPUTE = AGE SEX (c) SES (c);!only covariates, no DV
NDATASETS = 20; SAVE = missimp*.dat;
1) is the syntax correct in order to handle the two level structure ?
2) should TIME be included on "within =" ? Or should I remove it from the USEVARIABLES and add it to Auxiliary ?
3) Is it correct that I don't have to worry about missing on the DV's since they will be handeled through FIML.
You are now in the wide format with your data so multilevel modeling is not needed. You don't need the CLUSTER, BETWEEN, or WITHIN options. See Chapter 6 where the wide versus long format is discussed. See also Example 11.5 for imputation.
Thank you very much for your answer. My data is in the long format. I wanted to stay in the two level framework (and not move into LGCA using the wide format). My concern is that the imputations of level-2 covariates will not vary within subjets.
I am trying to create 10 imputed datasets for a twolevel model, however, I don't get any output file. When I run the input file, it seems to be working, creates a list file, but no datasets are created and the list file is empty.
My syntax reads: VARIABLE: NAMES ...; USEVARIABLES ARE zGender AtHome InSchool NoArrest Language100 zRace zCaseProgressReview Txlengt zTherapistID cohort; AUXILIARY = famav19 therexp teamyexp team1-team26; CLUSTER = zTherapistID cohort; CATEGORICAL = athome inschool noarrest; MISSING ARE ALL (999);
DATA IMPUTATION: IMPUTE = zgender(c) Language100(c); NDATASETS = 10; SAVE = TotalFile19.ftl*.dat
1) Would you have any idea what is going wrong? 2) Is this the correct way the specify a H1 model? 3)Is it true that the default is to estimate a variance covariance model? How would I need to specify a sequential regression model?