Problem in testing mediation effect

[har_ter]

I have created a formative measure and I’m testing its effect on firm performance.

My measure consists of two managerial activities both consisting of efforts (AE & RE) and design (AD & RD). In other words, the measure approximates the degree (strength) of the two activities and also the design (quality) of the two activities. All four measures are formative.

The measure has been validated by the guidelines of Diamantopoulos and Winklhofer (2001). The multicollinear indicators have been dropped.

However, the paths between efforts and design in both activities are very large around 0.66. This is theoretically meaningful as the efforts and design approximate different aspects of the same construct.

My problem is following. When I draw a model with just efforts AE and RE (independent variables) to firm performance (dependent variable) the paths are significant and the path coefficients are around 0.200-0.300.

The next step is to examine whether the design (AD & RD) of the two activities mediate the above tested paths to firm performance. In many cases the design mediates successfully the AE’s and RE’s performance. However, in several cases the introduction of design (AD & RD) into the model turns both the effort and design non-significant and also the path coefficients turn very close to zero. Testing of moderation is not reasonable as the desing and efforts are closely connected.

The question is:

1) Do the strong paths between respective efforts and design explain the above problematic results? (before significant paths (AE, RE) become close to zero and turn into non-significant, design (AD, RD) is also close to zero non-significant)?

2) Can I just draw the significant paths into the model (i.e. efforts and the successfully mediating designs) and leave the non-significant constructs out of the model (i.e. designs that are non-significant and which make the before significant efforts nonsignificant)? Doing this will lead to logical results – also the paths and R-squares are here highest. (These results are identical to second-order conceptualization of this phenomenon but have stronger paths and r-square).

Best regards: Harri

[Hengkov]

Hi,

Used four step.
Baron and Kenny (1986), Judd and Kenny (1981), and James and Brett (1984) discussed four steps in establishing mediation:

Step 1: Show that the causal variable is correlated with the outcome. Use Y as the criterion variable in a regression equation and X as a predictor (estimate and test path c in the above figure). This step establishes that there is an effect that may be mediated.

Step 2: Show that the causal variable is correlated with the mediator. Use M as the criterion variable in the regression equation and X as a predictor (estimate and test path a). This step essentially involves treating the mediator as if it were an outcome variable.

Step 3: Show that the mediator affects the outcome variable. Use Y as the criterion variable in a regression equation and X and M as predictors (estimate and test path b). It is not sufficient just to correlate the mediator with the outcome; the mediator and the outcome may be correlated because they are both caused by the causal variable X. Thus, the causal variable must be controlled in establishing the effect of the mediator on the outcome.

Step 4: To establish that M completely mediates the X-Y relationship, the effect of X on Y controlling for M (path c') should be zero (see discussion below on significance testing). The effects in both Steps 3 and 4 are estimated in the same equation.

Regards,

[bwilson]

Hi,

Used four step.
Baron and Kenny (1986), Judd and Kenny (1981), and James and Brett (1984) discussed four steps in establishing mediation:

Regards,

The work of Hayes is also good with his website.
PLS is superior to Sobel test for testing mediation as you can take advantage of bootstrapping capabilities.

regards
Brad.