Invariant results in measurement erro... PreviousNext
Mplus Discussion > Structural Equation Modeling >
 Suyin Chang posted on Saturday, January 18, 2014 - 4:30 am
Hi, I come to ask a question about the single-indicator reliability correction for measurement error.

As I understand, the way to code a single-indicator latent variable to account for measurement error in Mplus is:

latent by observed@1;
Where theta is equal var(observed) times 1-reliability (measured in percentage from 0 to 1).

Now, in the models I am working with, it doesn't matter how low or how high I input the reliability, not a single result in the model is affected besides the standardized coefficients and their SEs. Not the non-standardized coefficients, nor their SEs, nor the fit measures and their p-values, really nothing. And I really mean it: not a thing, even by decimals.

My code is precisely this:

MODEL: X1@2.8;
L1 by X1@1;
Y1 on X1 X2 X3;

Where var(X1) = 3.4

All results from the above specification are identical to:

MODEL: X1@0;
L1 by X1@1;
Y1 on X1 X2 X3;

With the exception of the standardized results (coefficients and p-values).

Is it something that can sometimes happen? Is there any known reason for such behavior?

Thanks for any tips you may have, and for your time.
 Bengt O. Muthen posted on Saturday, January 18, 2014 - 9:19 pm
I think you want to say

Y1 ON L1;

instead of

Y1 ON x1;
 Suyin Chang posted on Saturday, January 18, 2014 - 11:44 pm
Dear Dr. Muthen, yes it was an unfortunate typo. I meant Y1 on L1 (of course I double-checked to see if there was any similar typo in the code, but there isn't).
Back to top
Add Your Message Here
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Options: Enable HTML code in message
Automatically activate URLs in message