Hi quick clarification question, if two slopes are both negative, and positively correlate in a longitudinal parallel-process model, what exactly does that mean....both slopes are negative but positively correlate? Thanks! James.
I am interested in the correlation between the slopes of two variables. Because I was wondering whether the results would be the same when shared variance between the two variables were removed, I added the within time-correlations at each time point in the model specification. I have three questions hereby:
1. Was this the correct way to control for shared variance? 2. What are the pro's and contra's of controlling for shared variance? 3. The correlation between the slopes remains similar, but this result is no longer significant (despite the strong negative correlation). How should I interpret this?
It soundslike you are talking about the residual covariance at the same time point between two outcomes that are part of two different growth model. Is this correct? Is this residual covariance what you mean by shared variance?
jpmv posted on Wednesday, April 20, 2011 - 12:07 pm
I think allowing for contemporaneous residual covariance between outcomes of two different growth processes is often necessary to capture the effects of left-out time-varying covariates that influence both processes. This then avoids channeling too much of the correlation between the two outcomes through the growth factors. That is, the correlation you get between the growth factors is more trustworthy when you include these contemporaneous residual covariances.