Hi. I am doing growth modeling with 3126 paritipants. Working memory was collected over 8 data points. Before I add SES into the model, I set up a basic model. But the basic model does not have good model fit. Based on my attached output, is there anything you would suggest?
As you can see, I already used modification indexes to create relationships.
Thanks for all of your help!
Zack,
I would suggest a few things:
1. There appears to be a mistake in the factor loadings of the slope factor. The loading fore WM7 should be fixed to 6, and the one for WM8 should be fixed to 7, right?
2. As you have eight observations, you could try adding a quadratic slope in order to model more complex growth patterns.
3. Alternatively, you could try freely estimating the slope loadings (except for the first two).
4. The residual covariances estimated in the model seem rather arbitrary. Those should be based on plausible assumptions. For instance, you could assume that only residuals from adjacent time points are allowed to covary.
5. What is "WorkMemK"? Are there variables in the model beyond those shown in the figure?
Heiko
...what about growth mixture models. Sometimes the assumption that there is only one slope/intercept that explains the data adequately is not reasonable.
Mark
Dear Zack
first your model need to be rotated, as the Slope and ICEPT place in left of paper. if you see some bias I recommend you to look back your data. sometimes there is some outliers or huge covariance exist there. then you may check your outer loadings to be sure that the items are not highly correlated.
hope this help you.
As suggested by Mark, extending the model to a growth mixture model may be interesting, as you probably have a good sample size for that.
- Badges
- Science topic
- Similar topics
- Analytical Chemistry
- Extraction