| Message/Author |
|
|
|
|
Dear Drs. Muthen,
I am trying to plot my latent growth model. I have found most of the parameters in the Mplus output to plug into the multilevel model equation. However, I'm wondering about the residual variance of the Intercept. Does that go into calculating the Intercept at Level 2 (of a 2-level multilevel model or a single level growth model in Mplus)?
Specifically, I have four predictors and their unstandardized regression coefficients; I have intercept of the intercept; but, then, I have an r sub 0i that I think might be the residual variance of the intercept. Does that sound correct?
Thanks,
Mary |
|
|
|
|
| For the plot, you need the model-estimated means - or probabilities if the outcomes are categorical. With continuous outcomes, the means don't involve the residual variance of the intercept growth factor but with categorical outcomes they do. |
|
|
|
|
Thank you for the explanation. I have an outcome which is a count, modeled with a negative binomial distribution in a growth model with 3 time points. It looks like the output is in logodds, which I exponentiated to get sample means. So, it sounds like I do not use the residual variance of the intercept growth factor in the equation, right?
Mary |
|
|
|
|
| Actually, for counts just like for categorical DVs, these (residual) variances make a big difference. To get the estimated means of the count outcome, numerical integration using these variances is needed. That's hard to do on your own. Mplus provides the estimates in the plots. |
|
|
|
|
Thank you, that's very helpful. I will look at the plots.
Mary |
|
|
|
|
Hi Dr. Muthen,
Can you please tell me which of the plot options that I should select to get the estimates in the plots? I re-ran the input including Plot: and Type = plot3; but I only get four options for plots and I have looked through them but can't seem to find anything helpful. Can you please advise?
Thanks,
Mary |
|
|
|
|
| You should be using the SERIES option in the PLOT command. |
|
|
| Back to top |
