Dual Strategy for Detection of Confounding

Note: There is no hypothesis test for confounding. The appraisal of confouding is based on judgement!

Strategy A: Explore Preconditions for Confounding

    E/C relationship
    C/D relationship

    Computation:
> TABLES E C
> TABLES C D

If E & C are associated, and C is an independent risk factor for D, the potential for confounding is strong.

Strategy B: Sensitivity Analysis

Compute adjusted measure of association (aMA^ ; e.g., Mantel-Haenszel summary measure of association) and compare to unadjusted (crude) measure of association ( cMA^)

Computation:
> TABLES E D C

Then, compare aMA^ to cMA^ -- a meaningful difference suggests a strong potential for confounding.