|
|
 |
| Comparing models by looking at varian... |
 
|
| Message/Author |
|
|
| Anh Hua posted on Friday, August 21, 2020 - 1:48 pm
|
|
|
Dear Dr. Muthen,
I've run 3 models. Model 1: piecewise with 1 intercept and 3 slopes of reading on assessment #1; Model 2: where the intercept and 3 slopes (1 slope for each year of project) on reading assessment #1 predict students' reading performance on assessment #2; Model 3: where the intercept and 3 slopes on reading assessment #1 PLUS students' time varying covariate, which is learning environment they experienced in each year of the project, altogether predict reading assessment #2.
I want to compare Model 2 and Model 3 to determine if adding the time varying covariate of learning environment account for the variability of students' scores on assessment #2 better than just the intercept and the slopes alone. How do I get mplus to give me some type of additional R squared for this? I'd appreciate any advice you can give me. If this can't be done in mplus, I wonder if there's another way to compare the models and provide some meaningful context for the readers.
I'm attaching my current syntax of model 3 for your review. Model 2 has everything except for the last line, which was "g4_scaled_score ON di_mean_g2 di_mean_g3 di_mean_g4;". |
|
|
| Anh Hua posted on Friday, August 21, 2020 - 1:49 pm
|
|
|
I apologize for posting my syntax in a second post:
USEVARIABLES= score_1 score_2 score_3 score_4 score_5 score_6 score_7 score_8
score_9 months_1 months_2 months_3 months_4 months_5 months_6
months_7 months_8 months_9 g4_scaled_score di_mean_g2
di_mean_g3 di_mean_g4;
MISSING=.;
TSCORES = months_1 months_2 months_3 months_4 months_5 months_6 months_7 months_8
months_9;
ANALYSIS: TYPE = RANDOM;
ESTIMATOR=ML; !MLF was recommended by mplus software when I added parcc scaled scores
miter=50000;
mconv=0.000001
MODEL: i s1 | score_1-score_3 AT months_1 - months_3;
i s2 | score_4-score_6 AT months_4 - months_6;
i s3 | score_7-score_9 AT months_7- months_9;
g4_scaled_score ON i s1 s2 s3;
g4_scaled_score ON di_mean_g2 di_mean_g3 di_mean_g4; |
|
|
|
|
| I don't know about R-square improvements; I typically don't use that. Why not instead see if the time-varying covariates have significant effects when they get added to the model? You can do that via Model Test, testing if all of them are zero. And checking that the growth factors still have significant effects when adding the time-varying covariates. |
|
|
| Anh Hua posted on Tuesday, August 25, 2020 - 12:35 pm
|
|
|
Got it! Thank you Dr. Muthen, as always, for your advice.
Anh |
|
|
| Back to top |
|
|
