Mover Stayer Single Variable Model
Message/Author
 Chris Weber posted on Thursday, December 05, 2013 - 9:44 am
I am estimating a simple mover stayer model with a three category variable measured at five time points. Since the latent classes for each of the five waves is measured with a single variable, I constrain the item thresholds. Yet, I get the following error for each class:
*** ERROR in MODEL command
One or more pairs of ordered thresholds are not increasing in Class 1.
Check your starting values. Problem with the following pairs:
PID_1\$1 (15.000) and PID_1\$2 (-15.000)
PID_2\$1 (15.000) and PID_2\$2 (-15.000)
.....
Any tips on how to do this?

Clases are specified as follows:
classes= c(2) c1(3) c2(3) c3(3) c4(3) c5(3); !c is mover stayer

At the end of the file, I constrain the thresholds for each of the three classes within the five latent variables (t1-t5). This generates the error. For instance:

model c1:
%c1#1%
[pid_1\$1@15];
[pid_1\$2@-15];

%c1#2%
[pid_1\$1@-15];
[pid_1\$2@15];

%c1#3% !baseline
[pid_1\$1@-15];
[pid_1\$2@-15];

model c2:
%c2#1%
[pid_2\$1@15];
[pid_2\$2@-15];

....and so forth.
Thanks.
 Bengt O. Muthen posted on Thursday, December 05, 2013 - 10:05 am
You have to order the threshold values from low to high. So

[pid_1\$1@15];
[pid_1\$2@20];

corresponds to falling in the lowest of the 3 categories,

[pid_1\$1@-15];
[pid_1\$2@15];

corresponds to falling in the middle category, and

[pid_1\$1@-20];
[pid_1\$2@-15];

corresponds to falling in the highest of the 3 categories. Note that the threshold values are always increasing.
 Justin Delahunty posted on Wednesday, August 06, 2014 - 10:05 am
Dear Drs. Muthen,

I am trying to estimate a LTA mover-stayer model as per example 8.15. I had the same issue as above when attempting to constrain for equality (ordered thresholds not increasing), and was able to run the model by changing:

%c1\$1%
[c1_est\$1] (1)

to:

%c1\$1%
[c1_est\$1@-15] (1)

Is this the correct way to specify the order of the threshold values? c1_est contains 5 classes: is it acceptable to specify only the first threshold, or do I need to specify the threshold for each class?

Thank you,

Justin
 Bengt O. Muthen posted on Wednesday, August 06, 2014 - 2:23 pm
You are using @, so fixing the threshold, which means that the equality constraint does not take effect. With 5 categories you could instead say for example

[c1_est\$1*-.2] (1);
[c1_est\$2*-.1] (2);
[c1_est\$3*.1] (3);
[c1_est\$4*.2] (4);