LGM and within person correlations
Message/Author
 Paul Tremblay posted on Saturday, July 17, 2004 - 5:12 am
I'm designing a weekly diary study to assess alcohol consumption and severity of conflict situations with approx 28 time points. I'm interested in capturing the between as well as within person correlations between these two variables. I want to use LGM but I am not expecting any meaningful trajectory across time point but variation and significant within person correlations. I understand how this would be implemented in HLM (although not using latent variables) using Random Coefficient Regression model with Level 1 showing the within person correlations but would like to use a latent procedure. Can you tell me which model would be appropriate in Mplus. Thank you.
 Linda K. Muthen posted on Saturday, July 17, 2004 - 9:09 am
You can do it the way that HLM does it in Mplus. But if you want to use latent variables, you would estimate an intercept only model for each outcome. The variances of the intercept growth factors would capture between variance and individual variance would be captured in the residual variances of the outcomes.
 Paul Tremblay posted on Saturday, July 17, 2004 - 9:53 am
Would I regress one intercept growth factor on the other. I assume this would give me the between individual correlation but I'm not sure where I would get the within individual correlations (would it be by correlating the residuals for the observed variables at each time point?)Thanks.
 Linda K. Muthen posted on Saturday, July 17, 2004 - 10:49 am
You should correlate the two intercept growth factors. This is the default if they are exogenous variables.
 Paul Tremblay posted on Thursday, January 11, 2007 - 1:56 pm
I asked a while back about the possibility of calculating within-individual correlations in MPLUS. I am conducting some LGMs of weekly depression scores and weekly alcohol consumption scores. So far I have 13 consecutive weeks. I also want to look at the relationships between depression and alcohol. I understand that I could do this using LGM for two parallel processes and investigate the correlation between the growth factors for depression and alcohol. Given that I have (weekly) diary data I would like to be able to look not only at the between individual correlation between depression and alcohol but also the within-individual correlations. I have read that these two types of correlations can differ substantially. I understand that this is standard in HLM but I would like to use Mplus if possible. Is there a procedure in MPLUS that would provide the mean (and variance) of the within-individual correlation between depression and alcohol. Thank you.
 Linda K. Muthen posted on Friday, January 12, 2007 - 9:42 am
I'm not sure exactly what you want. Can you either write out the HLM model for what you want or send an HLM output that shows what you want to support@statmodel.com.?