How to get predicted values out of a ... PreviousNext
Mplus Discussion > Growth Modeling of Longitudinal Data >
Message/Author
 Shige Song posted on Sunday, June 11, 2006 - 3:41 am
Dear Linda and Bengt,

I have count variable y at 5 time points. Now I want to estimate a intercept only growth model like this:
--------------------
...
Count = ns1989-ns2000;
Analysis:
Type = random missing;
Model:
ns | ns1989-ns2000@1;
...
--------------------

Now I want to extract the predicted values, is it possible? I am thinking about something like the predict command in Stata.

The truth is that I want to use this set of count variable as covariate in a model, but there are large amount of missing values; so I want to replace missing values with the predicted score.

Thanks!

Shige
 Bengt O. Muthen posted on Monday, June 12, 2006 - 2:42 pm
There is currently not an option to compute the predicted individual values (we'll add it to our list), although that can certainly be done outside Mplus. On the other hand, Mplus has the (better) facility of predicting from the intercept (without having to use the estimated score as an intermediate step), directly estimating the slope of the influence from the intercept to what's predicted.
 Shige Song posted on Monday, June 12, 2006 - 3:14 pm
Thanks, Bengt.

I was reading the Mplus manual (for fun), and I noticed that when I request graphics from Mplus ("estimated individual values"), there is an option of exporting the graph data. What will be exported, if not the estimated score?

Shige
 Linda K. Muthen posted on Monday, June 12, 2006 - 5:00 pm
Estimated individual values are not available for count variables.
 Shige Song posted on Monday, June 12, 2006 - 5:04 pm
I see. Thanks for the quick answer.

Shige
 Jahun Kim posted on Thursday, June 16, 2011 - 5:33 pm
Dear Linda,

I have a basic question about growth model predicted value.
When I ran growth model, I constrained the variance of the slope.... then.. I got model predicted value of intercept.

I'd like to use the predicted value of each case when I run subsequent analyses.

How the predicted value is different from the original baseline value of the variable?
How can I justify use of the predicted value instead of using original variable score?
What is the predicted value and how Mplus calculate it?

Thanks in advance...
 Linda K. Muthen posted on Friday, June 17, 2011 - 3:04 pm
Are you asking how the intercept growth factor values for each individual differ from their values on the outcome at the first time point?
 Jahun Kim posted on Thursday, June 23, 2011 - 10:18 pm
Dear Linda,

I must clarify my question.
What I asked was....

When I ran growth model, I used SAVE DATA commend to get FScore. Because I fixed slope at 0, I could use predicted intercept, which was found in SAVEDATA information "I-SE" (correct?) for subsequent analyses.

Could you let me know the equation of calculating this score of each case?

Thanks.
 Linda K. Muthen posted on Friday, June 24, 2011 - 10:56 pm
You receive a factor score for the intercept growth factor i. i-se is the standard error or the factor scores. See Technical Appendix 11 on the website to see how factor scores are estimated.
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: