|
|
 |
| Model with all single-indicator laten... |
 
|
| Message/Author |
|
|
| Dusida Dee posted on Thursday, July 29, 2010 - 2:46 pm
|
|
|
I am working on a path model with 8 seperate variables that seem to be all 'observed' e.g. mean scores from individual psychometric scales, GPA.
If I would like to make a full SEM model, that is to keep the measurement part of it instead of doing merely a path analysis, can I do measurment models for all the 8 variables all of which will have just one indicator each? I know there is a method for this which deals with fixing the variance parameters and stuff (sorry I don't know how to explain it technically). But I'm just concerned whether or not such a 'weird-looking' model is acceptable as I've never seen one.
Please kindly enlighten me.
Many thanks in advance from a student from Thailand. |
|
|
|
|
| You can do this. See slide 44 of the Topic 1 course handout to see how it is done. Note, however, if you reliability estimate is not extremely good, you can create a whole new set of problems. |
|
|
| Dusida Dee posted on Thursday, July 29, 2010 - 11:10 pm
|
|
|
| Thank you very much for your answer. |
|
|
| Back to top |
|
|
