Vincenzo,
just fix the factor loading of an indicator variable to one and the error variance to zero. Then, the manifest variable will be equal to the latent variable. Do you nee help with the syntax?
Kind regards, Karin
Dear Vincenzo,
Its related to your research question or hypothesis.If you want to get a model from your data, first you should check exploratory factor analysis with SPSS, then do a confirmatory factor analysis with LISREL.
Hi,
Is this method applicable to ordinal variables as well? My issues is with an ordinal dependent variable. If I try to define this ibserved variable as a dependent to an independent latent variable, it takes it as another indicator variable to that latent variable. I have tried to avoid this problem by creating a single (ordinal) indicator latent variable as the dependent variable. The model works, but is this the right approach?
Thanking you in advance
Sandeep
Good morning,
I guess I'm having same problem of Vincenzo. Lisrel is giving me back " The model does not converge". I am not sure whether I correctly specified the model for a unique dependent factor composed by a single item.
Can someome help me out?
This is the script I prepared:
da ni=11 no=103 ma=cm
cm=tot.cov fu
la ADP B1 B2 O1 O2 V1 V2 E1 E2 E3 E4
se 1 2 3 4 5 6 7 8 9 10 11 /
mo ny=1 nx=10 ne=1 nk=4 ga=fu,fi be=fu,fi lx=fu,fi ly=fu,fi
lk BENEFIT ORG VENDOR EXT
le ADOPT
PA LX 2(1 0 0 0) 2(0 1 0 0) 2(0 0 1 0) 4(0 0 0 1)
PA GA 1 1 1 1
pd ou ALL ad=OFF
Well, you could define the error variance of ADP to be zero:
MO ... TE=FU,FI
This should fix one of your problems.
However, I think that there are some other problems in your model. As is obvious, there are three latent predictor variables measured by two indicators each. For a well-defined model, you would need at least three indicators. Therefore, this may be the cause of your estimation problems.
In order to check whether this is the problem just fix all factor loadings to one (assumption of tau-equivalence) and estimate your model. If it converges, than measuring latent constructs by two indicators only was the problem.
HTH
Karin
Thank you very much for your answer Karin.
I solved the first issue as you mentioned. Unfortunately I am a beginner in LISREL and I do not actually know how to fix all the factors loading to 1 in order to toubleshoot my model. Would you mind helping me out on that?
Thank you very much in advance
Kind regards,
Riccardo
Riccardo,
after the line " PA LX 2(1 0 0 0) 2(0 1 0 0) 2(0 0 1 0) 4(0 0 0 1) " add the following two lines:
FI LX 1 1 LX 2 1 LX 3 2 LX 4 2 LX 5 3 LX 6 3
(This line sets six loadings to zero - do not forget the blanks between the values)
VA 1 LX 1 1 LX 2 1 LX 3 2 LX 4 2 LX 5 3 LX 6 3
(This line fixes all six loadings to one)
This should work, hopefully.
Kind regards, Karin
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Structural Equation Modeling