Known Reliability
Message/Author
 Lindsay Bell posted on Wednesday, March 19, 2014 - 2:00 pm
Hi -

I am estimating a model with all observed variables (nothing latent), but my dependent variables, along with one of the independent variables, have known reliabilities, which I would like to adjust the parameters for in the analysis. How can I do that?

Thanks,
Lindsay
 Linda K. Muthen posted on Wednesday, March 19, 2014 - 2:37 pm
See Slide 44 of the Topic 1 course handout on the website.
 Lindsay Bell posted on Sunday, March 23, 2014 - 4:49 pm
Thanks, Linda. I have looked at the slide but am concerned that I am misunderstanding something. My goal is to adjust the regression coefficients for the reliability of the variables, but what I have done has not changed the coefficients at all, just forced the residual variances to be equal to the values that I calculated.

From Slide 44, I used the syntax y@a, where a = (1-reliability)*sample variance. So, for example, the variable fallpair has a reliability of 0.81 and a sample variance of 0.896, so I calculated (1-0.81)*0.896=0.17024. A sample of my model syntax is below. Can you tell me what I'm doing wrong? Also, why do I no longer get chi-square and RMSEA values when I do this reliability adjustment?

Thank you,
Lindsay

fallpair@0.17024;
parisei@0.3387;
income@0.10724;

fallpair ON parented
parisei
income
gender
age
minority
;
 Linda K. Muthen posted on Sunday, March 23, 2014 - 6:32 pm
In Slide 44 it shows

In Mplus:

f BY y@1;
y@a;

where a =
&#952;

You need to put a factor behind each variable and use the factors in the ON statements, for example,

f1 BY fallpair;
fallpair@0.17024;
Post:
This is a private posting area. Only registered users and moderators may post messages here.