Syntax IRT
Message/Author
 jan mod posted on Sunday, July 23, 2017 - 3:31 pm
Dear professor,

Someone told me that the following syntax calculates equalities for common items of each wave of items (with imposed threshold invariance on the common items and three correlated theta variables).

1) Is this correct?

2) I could not find it in the user guide or a book. Is there somewhere an explanation?

DATA:
FILE = three.dat;

VARIABLE:
NAMES = a1-a172;
CATEGORICAL = a2-a172;
USEVARIABLES = a1-a172;
MISSING = ALL(99);
IDVARIABLE = a1;

ANALYSIS:
ESTIMATOR = ML;

MODEL:
theta1 by a60-a111*
a52-a59* (a1-a8);
theta2 by a2-a9* (a1-a8)
a10-a17* (a9-a16)
a18-a51*;
theta3 by a112-a119* (a9-a16)
a120-a172*;
theta1-theta3@1;

[a52\$1-a59\$1] (b1-b8);
[a2\$1-a9\$1] (b1-b8);
[a10\$1-a17\$1] (b9-b16);
[a112\$1-a119\$1] (b9-b16);

PLOT:
TYPE = PLOT1 PLOT2 PLOT3;

SAVEDATA:
SAVE = FSCORES;
FILE = three.dat;
 Bengt O. Muthen posted on Sunday, July 23, 2017 - 5:32 pm
Q1: yes

Q2: See UG ex 5.10 for threshold restrictions. Look at the multiple indicator papers we have posted under Papers, Growth Modeling. Your model is a simplified version of growth modeling (no growth factors). See also today's answer to Laura Johnson.
 jan mod posted on Sunday, July 23, 2017 - 6:09 pm
Thank you, so to make it extra clear to me: the syntax:

for each (a1-a8) and (a9-a16),
theta1 by a60-a111*
a52-a59* (a1-a8);
theta2 by a2-a9* (a1-a8)
a10-a17* (a9-a16)
a18-a51*;
theta3 by a112-a119* (a9-a16)
a120-a172*;

you get a corresponding threshold invariance like:

[a52\$1-a59\$1] (b1-b8);
[a2\$1-a9\$1] (b1-b8);
[a10\$1-a17\$1] (b9-b16);
[a112\$1-a119\$1] (b9-b16);
 Bengt O. Muthen posted on Monday, July 24, 2017 - 4:29 pm
Right.