Hi,
I have a problem with SEM using Lisrel. I am using a Lisrel for my SEM modelling. All my variables are ordinal. Hence the indicator variables for the independent latent variables (intention) and the observed dependent variable (behaviour) are all ordinal. Now if I try to define an observed variable as a dependent variable to the latent variable, it assumes it to be another indicator variable to the latent variable. In order to solve this problem, I have tried using a single indicator latent variable. Hence I create a latent variable where behaviour is a single indicator and this latent variable then becomes the dependent variable to the latent variable intention remains the independent variable. The model works but reading the above argument I have doubts on its reliability. So I have two questions:
1) In Lisrel, how can I treat an observed variable - behaviour, as a dependent variable to a latent variable without it being mistaken as another indicator variable of the independent latent variable intention
2) If this is not possible then can I use a single indicator latent variable to define my dependent variable
Thank you in advance
Sandeep
Hi Rimantas,
Thanks for pointing my mistake. Also thanks for your answer, this is very helpful. Also I read a few posts on this method by fixing the loading to 1 and error variance to 0, however they were in relation to a continuous indicator variable. In my case I am using an ordinal variable. Can I use the same method?
Thanking you in advance for your answer.
Sandeep
Thanks Rimantas and Charbel,
I am fine with the use of ML, however please advise if I can do so with my particular dependent variable. The dependent variable i.e. behavior is actually a binary (0,1) variable. I am using it as ordinal because my Lisrel student version gives me only two practical options of defining the variable as either ordinal or continuous. Under these circumstances, do you think it will be right to treat this variable as continuous (using ML)? All my other variables are ordinal (on a scale from 1 to 5).
Also is there a recent literature which supports this view - which I can use as a reference to support this view in my thesis?
Thanking you in advance for your support.
Sandeep
Dear Rimantas,
Thank you ... this greatly useful. I think I will give MPlus a shot, in the meanwhile will also go through this paper. I may have more doubts though, and would like to direct some more questions to you.
Thanking you again.
Sandeep
Dear Rimantas,
I went through the paper you suggested. Basically the authors have used a two step process. Although I haven't exactly followed how it was achieved. I have written to the authors for more clarity and wait for their feedback.
In the meanwhile I was going through release notes of Lisrel 9.2 - i.e. the version that I am using. It appears to me that they have added generalized linear models to its capabilities. I tried using SurveyGLIM in this version, however it does not seem to handle latent variable in a SEM format. I may be wrong however.
Would you have any clarity on whether this version can be used to define single indicator dichotomous variables - or deal with such variables through any other means.
Also the MPlus demo version allows too few variables to run my model - so that seems out of reach for now. I need to find a way to work with Lisrel 9.2
Also I am exploring a method on treating all the variables as ordinal (and not continuous). In which case I can use a single indicator ordinal variable by imputing polychoric correlation matrix and Asymptotic cavariance matrix as input in the SEM. Are you aware of this option? What would you advice?
Thanking you in advance
Regards
Sandeep
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis