Zero-inflated ordered probit
Message/Author
 Roger Brown posted on Wednesday, March 02, 2016 - 7:15 am
Can we build a zero-inflated ordered probit model in Mplus? If not, would a ZINB model be a close approximation?
 Bengt O. Muthen posted on Wednesday, March 02, 2016 - 2:59 pm
Are you thinking about more than 2 outcome categories? Is the outcome ordinal?
 Roger Brown posted on Wednesday, March 02, 2016 - 5:11 pm
Yes, 5 ordered categories
 Tihomir Asparouhov posted on Thursday, March 03, 2016 - 9:53 am
Zero-inflated modeling is equivalent to a two-class model with all zeros in one of the two classes. You can use zero-inflated ordered probit by introducing a latent class variable C(2) for each variable that has a zero inflated distribution.
 Nikola Zaharakis posted on Saturday, July 13, 2019 - 9:10 pm
Dear Professors,
I am trying to test a cross-lagged panel model with ZIOP. I have 2 timepoints and 2 outcomes. 1 is binary (drug use – yes, no); 1 is ordinal (drug communication – 0 times, 1 time, 2 times, 3 or more times). The ordinal variable is zero-inflated.

1. Does my syntax appear correct? I cannot get the model to converge, whether I run the simple model, or include additional covariates to provide more information for the model.
2. Is there a way I can include the autocorrelation between the outcomes? Loading the outcomes onto latent variables, in order to correlate the errors makes the model more complicated.

Syntax:
USEVARIABLES are Y3MJ01 Y4MJ01 Y3com Y4com male age;
missing = all (99);
CATEGORICAL = Y4com Y4MJ01;
classes = c (2)
ANALYSIS: TYPE=MIXTURE;
ESTIMATOR = MLR;
ALGORITHM = INTEGRATION;
INTEGRATION=MONTECARLO;
MODEL:
%overall%
[Y4com\$1];
Y4com on Y3com Y3MJ01 male age ;
Y4MJ01 on Y3com Y3MJ01 male age;
Y3com with Y3MJ01;
%C#1%
[Y4com\$1@0];
Y4com on Y3com Y3MJ01 male age ;
Y4MJ01 on Y3com Y3MJ01 male age;
Y3com with Y3MJ01;
%C#2%
[Y4com\$1];
Y4com on Y3com Y3MJ01 male age ;
Y4MJ01 on Y3com Y3MJ01 male age;
Y3com with Y3MJ01;
 Bengt O. Muthen posted on Sunday, July 14, 2019 - 11:45 am