|
|
 |
| One p-value difference between MPlus ... |
 
|
| Message/Author |
|
|
| Julie Dav posted on Tuesday, June 26, 2018 - 9:53 pm
|
|
|
Hello,
I am hoping you might be able to help me understand something. I have a serial mediation model:
m1 ON x a b c d;
m2 ON x m1 a b c d;
y ON x m1 m2 a b c d;
Where:
x is a binary variable (2 experimental conditions);
m1 and m2 are continuous mediators;
a b c d are covariates - all continuous except one which is binary;
y is a continuous outcome variable.
I initially used PROCESS macro for SPSS (with bootstrap=10000) to investigate direct and indirect effects. According to this method these are the significant relationships:
x --> m1 --> m2 --> y
m1 --> y
I switched to MPlus because this model is just the first step and I will need SEM for further analysis. What is confusing however is that when I tested the exact same model in MPlus using ML, bootstrap=10000, one of the relationships became nonsignificant although all the coefficients are pretty much the same. The p value for the m2 --> y part of the model is around .09 whereas in PROCESS it was .04. What could explain this one difference? |
|
|
|
|
| Perhaps your DVs have missing data which is better handled in the Mplus multivariate approach. If not, send your output to Support along with your license number. |
|
|
| Back to top |
|
|
