Dear researchers,
I am struggling to find information on partially latent models in AMOS (vers. 27). I have four factors in a model and a final outcome which is a manifest (ordinal) variable (because the factor was not sufficiently reliable). I have three questions:
1) Why does AMOS not provide significance values for this manifest variable? (It does not show up in the output under regression weights?)
2) Should I include a disturbance/error term on this variable as with endogenous factors or latent variables? (AMOS does not come with any warning)
3) Can observed variables only be used as exogenous or as controls?
I hope you understand my question. I would appreciate if you can direct me to a source or book where I can find more information on this issue. Thanks for the help in advance!
Martin Brygger Andersen, yes, you can use observed variables (or single-item questions) as exogenous/controls or/and endogenous variables in a SEM model. In your case, you have got an observed variable used as an endogenous/outcome variable. The disturbance/error term should be automatically included. One last issue is to decide whether you should treat this variable as an ordinal or a continuous variable. The former is technically considered the correct choice. Software like Stata, R, and Mplus allow for estimating models including an ordinal outcome variable, but I am not sure if AMOS does.
Thanks, I have found some information on it, and you are right. I don't know why AMOS does not include the observed endogenous variable in the output. Also, I don't believe I can treat the outcome variable as ordinal but I will try to look for information. However, I have a very large dataset (more than 100,000 observations). So I believe that I can still treat the outcome as ordinal without serious bias because of the sample size.
I realized I did a mistake (the output does not show the label but the variable name). I have been so used to analyzing factors that I expected the label to show up in the output instead. Quite embarrassing!
- Badges
- Science topic
- Similar topics
- Design
- Architecture