🚀 go-pugleaf

Julis Lane (jilane@AUVM.AMERICAN.EDU) asks.....

> When you omit a variable in a linear regression, the consequence is
> to bias the coefficient on included variables which are correlated
> with the omitted variable.  It's pretty straightforward to show
> the magnitude of this bias.
>
> Does anyone know how to show this in logistic regression?  Any help
> would be much appreciated.
>

##  As I see the problem, it's difficult to quantify the
    bias in logistic regression, although I must admit that
    it's not impossible. This problem can be viewed from
    the concept of confounding. If the omitted variable is
    indeed a confounder, then the change in regression
    coefficient (log-odds) from the model with and without
    the variable (in question) will be **meaningfully** different.

##  Does this give you any lead ? Have I understood your
    problem correctly ? Bombard me, otherwise !!!!

--  Cande V Ananth   (ananth@UNCMVS.OIT.UNC.EDU)
    Dept of Biostatistics, SPH
    Univ of North Carolina at Chapel Hill