Linear Spline Model
Message/Author
 Thomas A. Schmitt posted on Tuesday, February 16, 2010 - 2:25 pm
Hello,

Are there examples of fitting linear spline growth models in Mplus to nonlinear data. It may just be terminology, but I can't seem to find an example in the manual. Also, is there any rationale for fitting a linear spline growth model with only one intercept? Thanks!

Tom
 Bengt O. Muthen posted on Tuesday, February 16, 2010 - 2:29 pm
The closest thing to linear splines in Mplus would be "piecewise" growth modeling.
 Thomas A. Schmitt posted on Tuesday, February 16, 2010 - 6:08 pm
A paper I'm looking at claimed to have fit a linear spline in Mplus and, but it must be a piecewise growth model. Assuming this is in fact what they are doing, does it make any sense to fit the model with only one slope? To me, it seems that they would not be modeling the nonlinearity of the data if only fitting it with a single slope.
 Thomas A. Schmitt posted on Wednesday, February 17, 2010 - 11:10 am
I figured it out so you can ignore this question.
 Lisa M. Yarnell posted on Saturday, November 05, 2011 - 10:21 pm
Hello, I am fitting a spline -- or a piecewise model as it is referred to in the Mplus manual. I have four time points and two groups. Is there anything obviously wrong with the following code:

i s1 | YEAR1@1 YEAR2@2 YEAR3@2 YEAR4@2;
i s2 | YEAR1@0 YEAR2@0 YEAR3@1 YEAR4@2;

i with s1;
i with s2;
s1 with s2;

This setup did not work in Mplus, where it was suggested that model may not be identified. Is there some alternative way that I should be specifying this model to have the knot at Year 3?

Thanks!
 Linda K. Muthen posted on Sunday, November 06, 2011 - 10:16 am
The first piece is not identified with only two timepoints for both the intercept and slope growth factors. You can do an intercept only model or fix the variance of s1 to zero. Note that the time scores for the first piece should be 0 1 1 1 .