X -> Y: X negatively affecting Y.
Y -> Z: Y negatively affecting Z.
X-> Z: X positively effecting Z
Y as a mediator.
I have completed collecting data and upload to PLS-SEM. During analysis all the value of path coefficients became positive. Do I continue to the step or not?
Hello Acham,
The sign of a given relationship depends on how each variable in the target relationship is quantified.
For example: Among automobile models, higher powered cars are generally quicker than lower powered cars.
Case 1. Not a paradox, but a function of quantification:
Define power as amount of horsepower at the drive wheels/axle.
Define quickness as the highest km/h indicated on the speedometer after 10 sec of full acceleration on a clear and level road.
This relationship will be positive.
Case 2.
Define power as in case 1.
Define quickness as the elapsed time required to reach 100 km/h of speed, under full acceleration on a clear and level road.
This relationship will be negative. Both case 1 and case 2 demonstrate the proposition as stated.
In general, what could cause the sign of an observed relationship to be contradictory to expectation?
1. Quantification of one or both variables, as outlined above.
2. Characteristics of the specific population under investigation.
3. Measures of questionable technical quality.
4. The presumption of the direction of the relationship was based on flawed theory.
Good luck with your work.
X may effect Z in a positive or negative way as it appears to be the direct effect in your mediating model. The more relevant question is whether the effect of X on Z is significant with the mediator present. If it is insignificant then the sign is irrelevant as it overlaps with zero. This is the case of complete or full mediation as opposed to partial mediation, the latter occuring when the direct effect remains significant (and assuming your indirect is significant in both cases). Hope that helps provide some assistance.
You may also consider relabeling your terms by using M for your mediator instead of Y and instead use the label Y for your DV instead of Z. Further, in some disciplines Z is reserved for moderating variables such as covariates, control variables such as demographics like age. Obviously every discipline has different conventions for such labels so do check what the usual X, Y,Z,M may be reserved for in your own area perhaps by looking over some published works where mediation has been considered.
Regards Paul
- Badges
- Science topic
- Similar topics
- Phytochemistry
- Extracts