Comparing correlations with/out measu... PreviousNext
Mplus Discussion > Structural Equation Modeling >
Message/Author
 Laura Duncan  posted on Monday, April 15, 2019 - 1:58 pm
We are attempting to show what happens to the association between two approaches to classifying ADHD (one based on a diagnostic interview (present/absent) and the other based on a dimensional scale score converted to a binary classification at a specific threshold) when you go from a simple bivariate correlation between two indicator variables to a correlation based on measures comprised of repeated classifications. Is it meaningful to compare these associations (i.e., claim that the stronger association is attributable to reduced measurement error?

We have a figure that shows the model but it is an image file, we could share this by email as I am unable to post it here. I tried to recreate it in text format but the spacing went all over the place so I have removed it.
 Bengt O. Muthen posted on Monday, April 15, 2019 - 4:21 pm
I would have to know what kind of modeling you are doing when you say

"a correlation based on measures comprised of repeated classifications"
 Laura Duncan  posted on Tuesday, April 16, 2019 - 7:16 am
It's not modelling in the traditional sense. It is descriptive comparison of two correlation coefficients: one based on two binary variables x and y; the other based on two 'latent' variables x and y derived from repeated classifications of x at time 1 and time2 (x1 and x2) and repeated classifications of y at time 1 and time 2 (y1 and y2). The objective is to demonstrate what we already know, that controlling for measurement error will increase the strength of association between the two variables. Our concern is that the tetrachoric correlation generated in the first instance (x, y) is not 'comparable' to the standardized correlation generated in the second instance.
 Bengt O. Muthen posted on Tuesday, April 16, 2019 - 5:45 pm
This question is suited for SEMNET.
Back to top
Add Your Message Here
Post:
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Password:
Options: Enable HTML code in message
Automatically activate URLs in message
Action: