Test of Association, Case-Control Studies

Test of Association, Case-Control Studies (2-by-2 Tables)

(A) Set up null and alternative hypotheses (e.g., H₀: OR = 1 vs. H₁: OR not = 1)

(B) Set alpha level

(C) Test statistic is Chi-square(1, N = xx) = xx.xx, p = probability region in right tail of Chi-square with 1 degree of freedom

Uncorrected, Mantel-Haenszel, and Yates Corrected chi-square are computed by the TABLES <exposure> <disease> command. Yate's corrected method produces most conservative p value. Use EpiTable or Statcalc to compute chi-square values for cross-tabulated data. Fisher's test is recommended when an expected value in the table is less than 5.

(D) Conclusion: Reject H₀ if p < alpha