Good afternoon everyone,

For my master's thesis, I investigate the effect of four different job crafting methods (4 times X variable) on daily intrinsic motivation (Y) via daily job enrichment (M). For this, I use the method of Baron and Kenny, followed by a Sobel test and a bootstrap.

Initially, I test the four mediating conditions by making one model, in which I include the four crafting variables. This shows that one of the four crafting variables (relational job crafting) satisfies the four conditions. I conclude partial mediation. However, the Sobel test and bootstrap indicate that the mediation effect is insignificant. When I perform the same exercise and only include one crafting variable each time, the Sobel test indicates a significant mediation effect for relational job crafting. The bootstrap still gives insignificance.

Would you advise me to carry out Baron and Kenny's 4 four steps four times in the analysis part of my thesis so that I can then at least demonstrate significance with the Sobel test for that one crafting variable, or is it from a statistical and academic point of view better to test all x variables in one model (and then I cannot conclude mediation for any variable)?

Furthermore, I was also wondering how I could prove the existence of a full mediation effect, as I assume that significane of path a, b and c and insignficance of path c' does not automatically mean you have a full mediation.

For your information, I use Stata 16, and I work with multilevel regression.

Thank you in advance.

Best regards,

Otto Rommens