Piecewise PreviousNext
Mplus Discussion > Growth Modeling of Longitudinal Data >
Message/Author
 Marianne SB posted on Thursday, November 07, 2013 - 1:45 am
We have measured depression at the ages of 13, 15, 18, 21, 23 and 30. We want to model two pieces, one from 13-18, and one from 21-30.

In most papers and examples, a piecewise growth curve is specificed like this:

i s1 | Dep13@0 Dep15@.2 Dep18@.5 Dep21@.5 Dep23@.5 Dep30@.5;
i s2 | Dep13@0 Dep15@0 Dep18@0 Dep21@.3 Dep23@.8 Dep30@1.2;

However, this model does not fit the data well. In stead, we have tested this model, which provides good fit indexes:

i s1 | Dep13@0 Dep15@.2 Dep18@.5 Dep21@.5 Dep23@.5 Dep30@.5;
i s2 | Dep13@0 Dep15@0 Dep18@0 Dep21@.8 Dep23@1 Dep30@1.7;

What is the difference between the interpretations of s2 in the two models, and is it correct to use the latter model?
 Linda K. Muthen posted on Thursday, November 07, 2013 - 9:12 am
The time scores should reflect the distance between the measures. They should be

0 1 1.5 1.5 1.5 1.5

and

0 0 0 1 2 3.5
 Marianne SB posted on Friday, November 08, 2013 - 1:15 am
Thanks for your reply! I am not sure if I understand it correctly though, probably because this is not how I am used to think about regular one-piece growth curves.

Is it correctly understood that the time scores should not reflect time cumulatively from age 13 (s1) and then from 18 (s2)? Should the time scores reflect number of years (1 time score = 2 years) from the previous measure?

How can one center the time axis at e.g. age 18 if time scores does not reflect the distance from the center?
 Linda K. Muthen posted on Friday, November 08, 2013 - 10:04 am
Sorry the time scores should be:

0 1 2.5 2.5 2.5 2.5

and

0 0 0 1 1 2/3 4

Let me describe how I got them for further clarification.

Piece 1:
13 15 18
0 2 5
0 1 2.5

Piece 2:
18 21 23 30
0 3 5 12
0 1 1.66 4

Time scores should always reflect the difference in time between the measurement occasions. It does not need to be the absolute difference. The time scores for the two slopes in a piecewise model are not related. Each piece has its own slope.

To center at 18 subtract 2.5 from the time scores for the first piece:

-2.5 -1.5 0 0 0 0
 Marianne SB posted on Sunday, November 10, 2013 - 10:42 am
Thanks again.

What is the difference between your specification (cfi = .897 rmsea = .07) and the following:

0 .2 .5 .5 .5 .5
0 0 0 .3 .5 1.2

Piece 1:
13 15 18
0 2 5
0 .2 .5

Piece 2:
18 21 23 30
0 3 5 12
0 .3 .5 1.2

(cfi = .917 rmsea = .063)

Should they not provide the same fit?
 Linda K. Muthen posted on Sunday, November 10, 2013 - 11:58 am
Please send the two outputs and your license number to support@statmodel.com.
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: