Chapter 4. Odds Ratio Methods for Unstratified Closed Cohort Data

Summary

This chapter presents odds ratio methods for analyzing unstratified data from a closed cohort study in which there are two or more exposure categories. One of the aims is to compare asymptotic unconditional, exact conditional, and asymptotic conditional methods, and to demonstrate that in many applications the various methods produce similar numerical results. Recommendations are provided for choosing methods for particular situations.

The section and subsection headings of the chapter are as follows:

• Asymptotic Unconditional Methods for a Single 2 × 2 Table

  • Point Estimate

  • Log-Odds Ratio Transformation

  • Confidence Interval

  • Pearson Test of Association

  • Wald and Likelihood Ratio Tests of Association

• Exact Conditional Methods for a Single 2 × 2 Table

  • Hypergeometric Distribution

  • Confidence Interval

  • Fisher's Exact Test

• Asymptotic Conditional Methods for a Single 2 × 2 Table

  • Point Estimate

  • Confidence Interval

  • Mantel–Haenszel Test of Association

• Cornfield's Approximation to the Hypergeometric Distribution

• Summary of 2 × 2 Examples and Recommendations

• Asymptotic Methods for a Single 2 × I Table

  • Point Estimates, Confidence Intervals, and Pearson and Mantel–Haenszel Tests of Association

  • Test for Linear Trend