small coefficients in mediation analysis

[Hillel Alpert]

Does mediation analysis tend to produce a smaller coefficient for a variable in the full model than the coefficient for the variable in a corresponding non-mediation analysis, i.e., the same model without the mediation component? I observe this with logit regression models that I am running with and without using the KHB mediation method (Stata command khb)? If typical, what might be the statistical reason?

[Hillel Alpert]

I found the discrepancy was because the mediation analysis was actually a probit model while the non-mediaion analysis was a logit model. Resolved.