This page contains an example of a random effects growth model using both a latent trajectory model (LTM) and a hierarchical linear model (HLM). The data were primarily used as part of a symposium that was presented at the 1997 meeting of the Society for Research on Child Development (SRCD). Details about the symposium and participants can be accessed here, and a document detailing the data & measures can be accessed here.
The substantive area of interest was the relation between developmental trajectories of aggressive behavior and reading ability in children over time, and the relation between individual differences in these trajectories and several time specific child and family characteristics. The sample consisted of N=405 children drawn from the Children of the National Longitudinal Survey of Youth, about half of which were missing one or more of the repeated measures on aggression or reading ability. The measures of interest included four repeated measures of aggressive behavior taken at two-year intervals, four repeated measures of reading achievement taken at two-year intervals, Time 1 measures of cognitive stimulation and social support of the child in the home, and background measures of child gender, child age, and mother age.
The core data and SAS files are:
•The raw ascii data set for N=405 cases with partially missing data
•The SAS code that reads in the full data set & reads out data for other analysis
A subset of data from the complete N=405 was extracted for the example. This data is based on N=221 subjects with complete data at all four time points and contains four repeated measures of antisocial behavior, for repeated measures of reading ability, and a time 1 measure of child gender and cognitive support of the child in the home.
•the order of variables are: anti1-anti4 read1-read4 gen homecog subjid
•The sample data set of N=221 cases in raw ascii format (save to disk as 'antiread.dat')
•The sample data set of N=221 cases in CSV format (save to disk as 'antiread.csv')
•The sample data set of N=221 cases in HLM5 format (save to disk as 'antiread.ssm')
For the example, two growth models were estimated, one unconditional and one conditional. First, an unconditional linear growth model was estimated for the four repeated measures of antisocial behavior over time. Next, this linear model was regressed on two correlated predictor variables, gender and cognitive support. The models were estimated in six different SEM packages and in two HLM packages. These should all provide the same results within rounding error.
•A PDF with the path diagrams, Level 1 & 2 equations, and summary of results is here.
• The unconditional linear growth model for the t=4 antisocial measures:
•Amos (save file to disk,rename file to end in 'amw', & open directly in Amos)
--Resulting Amos text output
•CALIS (paste directly into SAS)
--Resulting CALIS text output
•EQS (paste directly into EQS)
--Resulting EQS text output
•HLM5 (paste directly into HLM5)
--Resulting HLM text output
•LISREL (paste directly into LISREL)
--Resulting LISREL text output
•MIXED (paste directly into SAS)
--Resulting MIXED text output
•Mplus (paste directly into Mplus)
--Resulting Mplus text output
•Mx (paste directly into Mx)
--Resulting Mx text output
•Alternative Mx script (courtesy of Mike Neale)
--Alternative Mx text output
•SIMPLIS (paste directly into LISREL)
--Resulting SIMPLIS text output
•The conditional linear growth model with gender & cognitive support as predictors:
•Amos (save file to disk,rename file to end in 'amw', & open directly in Amos)
--Resulting Amos text output
•CALIS (paste directly into SAS)
--Resulting CALIS text output
•EQS (paste directly into EQS)
--Resulting EQS text output
•HLM5 (paste directly into HLM5)
--Resulting HLM text output
•LISREL (paste directly into LISREL)
--Resulting LISREL text output
•MIXED (paste directly into SAS)
--Resulting MIXED text output
•Mplus (paste directly into Mplus)
--Resulting Mplus text output
•Mx (paste directly into Mx)
--Resulting Mx text output
•Alternative Mx script (courtesy of Mike Neale)
--Alternative Mx text output
•SIMPLIS (paste directly into LISREL)
--Resulting SIMPLIS text output
