Syntax for Cross-Lagged Panel Analysis PreviousNext
Mplus Discussion > Growth Modeling of Longitudinal Data >
Message/Author
 Colin Mahoney posted on Tuesday, February 12, 2019 - 8:14 am
Am I running the correct syntax to examine 2 variables at 3 time points in a cross-lagged panel analysis? Syntax is below:

VARIABLE:
NAMES are
PCL2 PCL3 PCL4 AC2 AC3 AC4;
MISSING is
PCL2(999) PCL3 (999) PCL4 (999) AC2 (999) AC3 (999) AC4 (999);

ANALYSIS: type=general;
estimator=mlm;

MODEL:

AC3 on AC2 PCL2;
PCL3 on PCL2 AC2;
AC4 on AC3 PCL3;
PCL4 on PCL3 AC3;

PCL2 with AC2;
PCL3 with AC3;
PCL4 with AC4;

OUTPUT:
stdyx
 Bengt O. Muthen posted on Tuesday, February 12, 2019 - 5:24 pm
That looks right. Also note the RI-CLPM approach shown on our website.
 Colin Mahoney posted on Wednesday, February 13, 2019 - 11:07 am
Thank you very much. I ran the RI-CLPM and the model did not converge.

With the syntax in my original post, my model fit was poor (see below). Any suggestions as to how to improve my RMSEA and my TLI? I expected the chi-square to be significant because this is a fairly large N, but I'm wondering about the rest. Thanks again.

MODEL FIT INFORMATION

Number of Free Parameters 23

Chi-Square Test of Model Fit

Value 150.883
Degrees of Freedom 4
P-Value 0.0000

RMSEA (Root Mean Square Error Of Approximation)

Estimate 0.158
90 Percent C.I. 0.137 0.180
Probability RMSEA <= .05 0.000

CFI/TLI

CFI 0.958
TLI 0.855

Chi-Square Test of Model Fit for the Baseline Model

Value 3550.314
Degrees of Freedom 14
P-Value 0.0000

SRMR (Standardized Root Mean Square Residual)

Value 0.036
 Bengt O. Muthen posted on Wednesday, February 13, 2019 - 11:20 am
I think your 4 df come from zero lag-2 paths. That is, the time 2 outcomes may predict the time 4 outcomes (directly).
 Colin Mahoney posted on Wednesday, February 13, 2019 - 11:43 am
So would I run it like this then?:

MODEL:
PCL3 on PCL2 AC2; PCL4 on PCL2 AC2;
AC3 on PCL2 AC2; AC4 ON AC2 PCL2;

PCL2 with AC2;
PCL4 with AC4;

When I run it this way, I get no fit indices. Should I have time 2 outcomes predict both time 3 and time 4 outcomes?
 Bengt O. Muthen posted on Wednesday, February 13, 2019 - 11:52 am
Correct. Perfect fit because the model doesn't have any zero paths. You can see which lag-2 effects are significant.

But you should try to get the RI-CLPM going because it may alleviate the need for lag-2 effects. You can send output to Support along with your license number.
 Noud Frielink posted on Saturday, April 06, 2019 - 8:03 am
Iím running a cross-lagged model with only two time points (interval 4 years). Several variables (a, b, c, d, and e), were measured at T1 and T2 with a sample size of 120. Given the two waves, I'm not able to use the RI-CLPM approach. Is the 'standard' cross-lagged panel analysis the best approach in this case, and if so, is there a tutorial available for helping to conduct this analysis?
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: