KHB method - controls

[Amber Drew]

Hi,

I am using the KHB method to test for mediation in logistic regression. I want to know if lived experiences with substance use (livedexp) mediates the relationship between the number of songs with substance use references known (drugsrap) and attitudes toward substance use (drugsB). My controls are age, gender, and race.

DV = attitudes toward substance use (binary), where 1 = positive attitudes, and 0 = negative attitudes
IV = # of songs with substance use references known (count), ranging from 0 to 14
MV = # of lived experiences with substance use (count), ranging from 1 to 7
CV = age, gender (Man vs Woman), race (Black vs Non-Black)

My first question is: where do the control variables go in the equation? Do they go before or after the double lines?

khb logit drugsB c.drugsrap || c.livedexp, summary

Second, before I test for mediation, do I need to run separate logit models testing the relationships between the IV and MV, and the MV and the DV? If I do this, how would I be able to compare the results of these to the khb output?

[Mike Lacy]

In the documentation for -khb-, they use the term "concomitant" for what you are describing as controls. Per the syntax diagram in -help khb-, then, you'll want something like:

khb logit drugsB c.drugsrap || c.livedexp, summary concomitant(age i.gender i.race)

The help file for -khb- is a bit thin, so you will likely want to check out the Stata Journal article listed in -help khb-.