dichotomous mediator with continuous IV and DV

[Protzko]

So I have a problem with mediator variables.  I have a dataset that has a continuous IV, a continuous DV but the mediator is dichotomous.  Also, the data is non-normal and refuses to get so when it comes to transforming.  Therefore, I am trying to find a robust method to run a regression on the data (bootstrapping, sobel, OLS, etc.) that can not only handle the non-normality, but can also deal with my dichotomous mediator.  Anyone have any ideas?

many thanks

~John

[Bruce Weaver]

Protzko wrote

So I have a problem with mediator variables.  I have a dataset that has a continuous IV, a continuous DV but the mediator is dichotomous.  Also, the data is non-normal and refuses to get so when it comes to transforming.  Therefore, I am trying to find a robust method to run a regression on the data (bootstrapping, sobel, OLS, etc.) that can not only handle the non-normality, but can also deal with my dichotomous mediator.  Anyone have any ideas?

many thanks

~John

John, you say "moderator" in the subject line, and "mediator" in the body of the message.  Which is it?  Regarding normality, bear in mind that it is the residuals from the model that should be approximately normal, but even that is not the most important requirement.  It is more important that the residuals be mutually independent, and not correlated with the explanatory variables (or with the fitted value of Y).

[Protzko]

Sorry about that, I changed the title; it is a categorical mediator variable.  The residuals are also all non-normal.  I would prefer to use a more robust regression technique than simply make an appeal to the CLT and go ahead.  As for the residuals, none are correlated with the explanatory variable,