Simulating LGM models PreviousNext
Mplus Discussion > Growth Modeling of Longitudinal Data >
Message/Author
 Jon Heron posted on Friday, November 11, 2011 - 12:26 am
Hi Bengt/Linda


I am simulating a linear growth model (3 repeat measures) with a single continuous
distal outcome regressed on both intercept and slope.

Having already specified "Y on I S;" I don't seem to be able to add a correlation between slope and the residual for my outcome (endogeneity?) - is this possible?

We are currently investigating instances when a two-stage model (regressing Y on fscores/BLUPS) causes bias hence I'm attempting to covary res(Y) with aspects of the growth model. As I type this, I'm wondering if I need to define my Y-residual explicity as a N(0,1) with a unit loading on the outcome.



many thanks as always


Jon
 Linda K. Muthen posted on Friday, November 11, 2011 - 11:42 am
That parameter is not identified. You could experiment by fixing it at different values.
 Jon Heron posted on Friday, November 18, 2011 - 12:34 am
Thanks Linda
 Jon Heron posted on Tuesday, November 22, 2011 - 9:40 am
Hi Bengt/Linda

as mentioned above, I have recently been simulating some simple LGM models.

We have been debating the pros and cons of allowing different parameters to vary across simulated datasets and have reached the conclusion that for our purposes the only aspect to vary would be the residuals for the repeated measures.

Noticing that in the Mplus manual (e.g. example 12.2) the majority of the parameters under MODEL POPULATION are merely freed at starting values, I set about fixing most of these in line with the model we wanted to simulate from. Note that I didn't change the fitted MODEL itself, merely the model I was simulating from.

This led to the discovery that the MODEL RESULTS appeared totally unchanged. Does it sound like I have done something wrong or is it not possible to fix parameters in this way?

many thanks, Jon
 Bengt O. Muthen posted on Tuesday, November 22, 2011 - 5:58 pm
Marking the parameters as fixed or free in Model Population command has no effect - only the values are used to generate data. It is in the Model command that fixed or free makes a difference because there they refer to the role of the parameter in the analysis (not the generation). I often copy the Model Population parameters and paste them into Model and therefore use in Model Population a combination of fixed and free as I would for the analysis.
 Jon Heron posted on Wednesday, November 23, 2011 - 1:19 am
Ah I see, thanks Bengt
Back to top
Add Your Message Here
Post:
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Password:
Options: Enable HTML code in message
Automatically activate URLs in message
Action: