Explained Variance
Message/Author
 learningmplus posted on Sunday, January 27, 2013 - 2:02 pm
Hi Linda and Benght,

I have a question regarding the proportion of explained variance in a mediation model.

If I posit partial mediation and the mediating path is not significant

I have independent variable to mediator; mediator to outcome; and independent variable to outcome.

The path from the independent variable to the outcome is significant; the path from the independent variable to the mediator is significant, but the path from the mediator to the outcome is not.

In this case, when speaking of the proportion of explained variance to i speak of the independent variable and the mediator contributing to X% of the variance in the outcome or is it only thr independent variable that contributes x% of the variance since the path from the independent variable to the outcome was significant but the path from the mediator to the outcome was not.

In other words, is the variance accounted for explained by the significant paths or by all the variables pointing to the outcome? it will be helpful if you can advise about this to include the correct information in my presentation.

When one has sequential mediators, is the proportion of explained variance in the dependent variable explained by the variable (mediator) that preceeds it or by all the variables in the model? I would appreciate your help with this, too.

Thanks very much.
 Linda K. Muthen posted on Monday, January 28, 2013 - 12:05 pm
In your model, the variance explained in the outcome is explained by both the mediator and the independent variable.
 learningmplus posted on Monday, January 28, 2013 - 1:29 pm
Thank you. I need to follow up briefly for clarification.

I have one independent variable, 3 mediators, and one outcome. Is this possible? I derived the model based on theory. Will the percentage of explained variance be accounted for by all 3 mediators and the independent variable?

Most importantly, is this the case even if, let's say one of the mediation paths is not significant? Or do we say that the proportion of explained variable is accounted for by X variable only if x to y is significant? In this explanation, I assume x is the mediator and x to y is the mediation path. I also have z and q as two additional mediators. z to y and q to y are significant but x to y is not. Do we say that z and q explain x% of the variance or do we say that x, z, and q account for % of explained variance. Thank you again.
 Linda K. Muthen posted on Monday, January 28, 2013 - 1:39 pm
Every variable on the right-hand side of an ON statement for an outcome explains the variance in the outcome whether they are significant or not. So if you have

y ON m1 m2 m3 x;

m1 m2 m3 ON x;

the variance in y is explained by m1, m3, m3, and x.