Hi,
I would need some help in understanding the SEM output. I run a latent change score analysis to test the change score of a variable of interest at two-time points (repeated measure) PLD1 and PLD2. in Mplus 8 Demo Editor.
This is the model
VARIABLE: NAMES ARE PLD1 PLD2;
ANALYSIS: TYPE=GENERAL;
ESTIMATOR=ML;
MODEL: !latent change score
LCS BY PLD2;
!Autoegression
PLD2 ON PLD1@1;
!Proportional change
LCS ON PLD1;
! variance
PLD1; PLD2@0;
!mean
[PLD2@0];
OUTPUT: TECH1; TECH4; MODINDICES (ALL);
SAVEDATA: FILE is newoutput.sav; SAVE=FSCORES;
FILE is lcsnew.dat;
This is the output I have
Number of Free Parameters 4
Loglikelihood
H0 Value 146.426
H1 Value 163.076
Information Criteria Akaike (AIC) -284.853
Bayesian (BIC) -276.155
Sample-Size Adjusted BIC -288.746
(n* = (n + 2) / 24)
Chi-Square Test of Model Fit Value 33.299
Degrees of Freedom 1
P-Value 0.0000
RMSEA (Root Mean Square Error Of Approximation)
Estimate 0.705
90 Percent C.I. 0.512 0.920
Probability RMSEA
Hi,
please first consider this
https://www.researchgate.net/post/Netiquette_proposal_for_questions_on_structural_equation_modeling
Especially §1 and §8.
Best,
--Holger
Hi Francesca Penza,
First off, I would check out this great tutorial on latent change score modeling by Rogier Kievit (Article Developmental cognitive neuroscience using Latent Change Sco...
). I have found it invaluable for my own work. You might notice that you're missing one constraint in your model (LCS by PLD2@1). That should make your model just identified, so your model fits become moot.Model fits can be a challenge to understand when you're first starting out, I get that. I would suggest starting with some work by Hu and Bentler (e.g., https://psycnet.apa.org/record/1998-03102-001 and https://www.tandfonline.com/doi/abs/10.1080/10705519909540118?journalCode=hsem20).
Finally, the MODINDICES command can be helpful for initial exploratory work, but I try to stay away from it. It basically tells you additional relationships to consider that would improve your model fit, but it doesn't keep in mind the importance of model sensibility, theory, etc. You want to define model structures that are in line with a theoretically/empirically sound rationale. Sometimes adding in the "extra stuff" in MODINDICES doesn't really make sense for what you're trying to do, and you can't fairly justify adding in those additional parameters on the back end when you try to publish.
Hope this helps!
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Structural Equation Modeling