Cross-lagged panel analysis with bina... PreviousNext
Mplus Discussion > Categorical Data Modeling >
Message/Author
 Gaye Ildeniz posted on Wednesday, November 07, 2018 - 12:16 pm
Hi,

Can I use a cross-lagged panel design in Mplus with a mixture of binary and continuous variables treated as both exogenous and endogenous within the design?

Thanks.
 Bengt O. Muthen posted on Wednesday, November 07, 2018 - 5:28 pm
Yes. Use Estimator WLSMV or Bayes in which case the relationships involving the binary variables are linear among the underlying Y* continuous latent response variables.
 Gaye Ildeniz posted on Thursday, November 08, 2018 - 2:38 am
Thank you!

And my binary dependent variable can both be a latent variable/construct (with a measurement model) or on it's own as an observed variable. Is that correct?

Thanks again.
 Bengt O. Muthen posted on Friday, November 09, 2018 - 1:18 pm
Right.
 Gaye Ildeniz posted on Tuesday, November 20, 2018 - 8:20 am
Hi again,
Thank you very much for your responses.
I'm still trying to find the most appropriate design for my research questions as well as adjusting to the time constraints. So I have a few more questions:

1. If I have 9 (8 of them being psychological constructs, 1 of them being past behaviour) variables measured at Time 1 and only 1 out of 9 (the past behaviour) measured again in Time 2, how can I analyse the estimated effect of 8 psychological variables on the Time 2 past behaviour outcome, controlling for the Time 1 measure? Does this fall under a full latent SEM or there is another way to model this due to the 2 time points?

2. Is it common to have several variables (by several, I mean 9 variables) in a 2 wave cross-lagged design? Does the number of variables create any issues when running the model in Mplus?

Thank you very much for your help.
 Bengt O. Muthen posted on Tuesday, November 20, 2018 - 5:25 pm
1. Perhaps these variables are indicators of a latent variable (factor) in which case you can have measurement invariance over time as shown in the multiple indicator version of the RI-CLPM at

http://www.statmodel.com/RI-CLPM.shtml

2. Only if they are indicators of a factor, otherwise the analysis is typically done one variable at a time - I think.
 Nicole S posted on Sunday, February 03, 2019 - 11:50 pm
I’m working on a cross-lagged model with a binary variable (x1 & x2) and an observed continuous variable (y1 & y2) each measured at two separate time points using the default WLSMV. The output comes up with errors when including the “x1 WITH y1” statement. When I exclude that statement it runs fine, and the diagrammer appears to account for that relationship but doesn’t provide a point estimate for it. I’m sure it’s really obvious but I’m just wondering why that is?

ANALYSIS:

PROCESSORS = 4;

MODEL:

y2 ON x1
y1;

x2 ON y1
x1;

x2 WITH y2;
x1 WITH y1;

OUTPUT:

SAMPSTAT STDYX
 Bengt O. Muthen posted on Monday, February 04, 2019 - 1:21 pm
You should not include x1 with y1 because that is the IV part of the model (see today's answer to "jb"). You get those estimates from Sampstat.
 Nicole S posted on Tuesday, February 12, 2019 - 9:58 pm
Thanks for the reply

Sampstat will only give me the estimates for x2 with y2, not x1 with y1, do you know why this might be?
 Linda K. Muthen posted on Wednesday, February 13, 2019 - 6:44 am
Do a TYPE = BASIC; with no MODEL command to get these.
 Elena Zubielevitch posted on Tuesday, July 21, 2020 - 4:37 pm
I am running a cross-lagged model with one latent continuous variable and one binary observed variable with known links. I am trying to figure out the most appropriate estimator to use.

I understand from these forums that with WLSMV I would need to regress the latent variables onto the observed variables to account for their association. I have two questions:

1.Is regressing the exogenous latent and observed variables appropriate given it still specifies a direction between them?

2.If I use ML instead, can I account for their association by specifying a WITH statement between the latent and observed exogenous variables, or is this not appropriate?

3.I’ve noticed doing so inflates the sample size so I’m wondering why.

Sorry if these questions don’t make sense, I’m new to Mplus.
 Bengt O. Muthen posted on Wednesday, July 22, 2020 - 6:04 pm
CLPM has not yet been developed/explored for cases with categorical outcomes. It's in progress.
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: