|
|
 |
| Describing CFA with categorical varia... |
 
|
| Message/Author |
|
|
| Alex R posted on Thursday, August 30, 2018 - 5:44 pm
|
|
|
I recently redid some analyses in Mplus, and I'm trying to figure out how exactly to describe the difference between a measurement model with categorical variables as opposed to continuous variables. Without going into other details (which I can provide upon request), am I expressing myself correctly when I say the following:
"Due to the zero-inflated, non-normal distributions of the X observed variables, there were concerns that using the linear regression model for the measurement of the X tendency (as a latent variable) led to biased estimates of its relationship to other personality traits, and possibly misleading findings. Therefore, to confirm that the results were robust to violations of normality, CFA analyses were rerun using the logistic regression model, whereby the factor indicators of X were treated as ordered categorical variables rather than continuous variables." |
|
|
|
|
That's ok. At least if by
"the linear regression model for the measurement of the X tendency (as a latent variable)"
you mean the linear regressions for the factor indicators on the factor (a linear factor analysis model). |
|
|
| Back to top |
|
|
