No zero order correlations PreviousNext
Mplus Discussion > Structural Equation Modeling >
Message/Author
 JW posted on Monday, September 24, 2007 - 8:48 am
Hi. I'm investigating a moderated mediational model. In conducting the zero-order matrix between all model variables, there was no correlation between the predictor and two out of three hypothesized mediators, but there was a significant correlation between all mediators and the dependent variable. What are the implications for my model as this was contrary to my predictions! Do I just retain the hypothesized mediator that is correlated with both the IV and the DV in my subsequent model analysis and explain the other two were excluded?? Or is it possible that a relation may only be seen when the moderator is tested? Any help would be much appreciated.
 Bengt O. Muthen posted on Monday, September 24, 2007 - 6:24 pm
I would determine the mediators' importance through regression analyses using all predictors instead of focusing on correlations. And consider the Model Indirect results. But it may well be that you have to abandon two of your mediators.
 JW posted on Monday, October 01, 2007 - 8:11 am
Thanks. Could you explain why you'd conduct a regression analysis - most studies report zero-order corrolation results for assessing feasibility? This would help my understanding more. Plus, would you suggest a standard multiple regression - entering all predictors simultaneously.
 Bengt O. Muthen posted on Monday, October 01, 2007 - 7:00 pm
I assume you have several predictors of the mediators, one of which is the central predictor (such as treatment vs control). Regression analysis, entering all predictors simultaneously, is suitable when you want to study the influence of a key predictor controlling for other predictors.
 JW posted on Friday, October 05, 2007 - 11:14 am
actually, i just have one variable as my predictor, with one variable for each mediator and one variable as DV. More a path analysis set up. I entered all variables in a standard regression in predicting the DV, and found that the predictor does not add significant unique contribution (by t statistic), but the two mediators do.
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: