Formative construct, categorical indi...
Message/Author
 Mr Mo DANG-ARNOUX posted on Wednesday, August 29, 2012 - 1:17 am
Dear Pr Muthen's and Mplus experts,

I am wondering how to define a formative construct, say F, based on ordinal indicators, say X1-X4. This formative construct predicts two latent (reflective) factors, G1 and G2, defined by Y1-Y3 and Y4-Y6.

I assumed the model could be specified as:

CATEGORICAL X1-X4 Y1-Y6;

G1 BY Y1-Y3;
G2 BY Y4-Y6;

F ON X1@1 X2-X4;
F@0;

F BY G1* G2*;

However, Mplus issues errors like:
*** ERROR in VARIABLE command
CATEGORICAL option is used for dependent variables only.
X1 is not a dependent variable.

How should I address this? My ordinal indicators X1-X4 have only 3 to 4 categories and floor effects, so that I am reluctant treating them as continuous variables.

Thank you very much for you attention and any guidance,

Mo
 Linda K. Muthen posted on Wednesday, August 29, 2012 - 9:41 am
There is no categorical data methodology involved with formative indicators. Leave them off of the categorical list. Formative indicators are basically sums.
 Mr Mo DANG-ARNOUX posted on Wednesday, August 29, 2012 - 6:53 pm
Thank you very much for your as ever fast answer, Dr Muthen.
 Hey Jung Jun posted on Sunday, September 16, 2012 - 9:25 pm
I am a new user of mplus.
I have two questions.

1) Can we use categorical variables like dummy variables as formative indicators in the formative measurement model?

2) If formative indicators are not statistically significant, do we have to remove them from the model and leave only those that are significant indicators on the model?

Thanks a lot.
Heyjung
 Linda K. Muthen posted on Monday, September 17, 2012 - 8:02 am
1. Yes but do not put them on the CATEGORICAL list. They are treated as continuous variables.

2. No.
 Hey Jung Jun posted on Monday, September 17, 2012 - 3:59 pm
Thank you very much. Heyjung