Interpreting effects in Bayes ML path...
Message/Author
 Tahani Z. K. Fourah posted on Friday, October 05, 2018 - 12:43 am
Hello Drs Muthen!

Could you please help me in interpreting one issue related to my results of multilevel path analysis using BAYES estimator?

I examined two models: random intercept, random slope models.

X,Z and K are level-1 predictors
A, B, C and D are level-2 predictors.
Y is the outcome.

------Random intercept model----
%WITHIN%
Y ON X;
Y ON Z;
Y ON K;
Z ON K;

%BETWEEN%
Y ON A B C D;
A WITH B C D;
B WITH C D;
C WITH D;
-----------------------
In the above model, K was found to be a significant negative predictor of Z (slope of K = -.149, credibility interval [-.237,-.062]).

However, when we allowed for the slopes to vary across groups as in the below model, slope of K was not significant (slope of K = -.124, credibility interval [-.256,.010] while the variation of the slope across groups was significant ( variance of slope =.059, credibility interval [.014,.158]) .

------Random slope model----
%WITHIN%
s1 | Y ON X;
s2 | Y ON Z;
s3 | Y ON K;
s4 | Z ON k;

%BETWEEN%

Y ON A B C D;
A WITH B C D;
B WITH C D;
C WITH D;
s1 s2 s4 s3;
--------------------------------

So, how one can interpret the effect of K on Z in each model?
Thanks!
 Bengt O. Muthen posted on Saturday, October 06, 2018 - 6:10 am
If your slope variance is substantial, I think you need to go with that model version, that is, the slope mean is not significant.
 Tahani Z. K. Fourah posted on Saturday, October 06, 2018 - 1:18 pm