Conflicting Beta Signs

Skip to first unread message


Oct 30, 2019, 6:48:30 PM10/30/19
to lavaan
Hi all,

I just ran a simple (standard) mediation model, where A has a direct effect on B, with a mediator M.

[path c]: A predicts B - positive beta.
[path a]: A predicts M - positive beta.
[path b]: M predicts B - negative beta.

[indirect ab]: A-M-B - negative.

All are significant (p < 0.05).

What does this mean (i.e., conflicting beta signs), and how should I go about interpreting this mediation model?

Thank you in advance!

Gavin Brown

Oct 30, 2019, 7:32:37 PM10/30/19
to lavaan
I understand this kind of result as possible non-transitivity in results. Non-transitivity occurs when positively correlated variables have inverse relations to a 3rd variable. Of course you want to check the confidence intervals to be sure the values are truly different to zero. But when the correlation between the variables is not high it is possible to discover non-transitivity. This type of phenomenon is quite feasible as long as the correlations are sufficiently different from 1 (Langford, Schwertman, & Owens, 2001). However, structural equation approaches can
help deal with cases where the correlations among predictors are large (Maruyama, 1998). 
Langford, E., Schwertman, N., & Owens, M. (2001). Is the property of being positively correlated transitive? The American Statistician, 55(4), 322–325. doi:10.1198/000313001753272286
Maruyama, G. M. (1998). Basics of Structural Equation Modeling. Thousand Oaks, CA: Sage.

Edward Rigdon

Oct 30, 2019, 9:34:59 PM10/30/19
What's the problem?
Harsh supervision (X) speeds task completion (Y)
Harsh supervision (X) encourages resistance (M)
Resistance (M) slows task completion (Y)
Positive direct effect, negative indirect effect.

You received this message because you are subscribed to the Google Groups "lavaan" group.
To unsubscribe from this group and stop receiving emails from it, send an email to
To view this discussion on the web visit

Terrence Jorgensen

Nov 2, 2019, 11:57:37 AM11/2/19
to lavaan
I agree with Ed, there is not apparent problem.  This is related to suppression effects, which many also (incorrectly) see as a contradiction.

Terrence D. Jorgensen
Assistant Professor, Methods and Statistics
Research Institute for Child Development and Education, the University of Amsterdam
Reply all
Reply to author
0 new messages