A Five-Region Hypothesis Test for Exposure-Disease Associations

Abstract

Characterizing exposure-disease associations is a central issue in epidemiology, one which epidemiologists often approach by adopting the index of the odds ratio and presenting its point estimate, p-value and confidence interval. In this study, the parameter space of the odds ratio is partitioned into five mutually exclusive regions corresponding to ‘strong protective factor’, ‘weak protective factor’, ‘no association’, ‘weak risk factor’, and ‘strong risk factor’, respectively. The authors presented a suite of statistical methods tailored to such a five-region demarcation, including methods for hypothesis testing, confidence interval estimation and calculation of the sample size needed to obtain the desired level of statistical power. The authors show that the five-region methods can efficiently and informatively describe a putative exposure-disease association, including its presence or absence, as well as its direction and strength (if any association exists). Three published results were re-analyzed to demonstrate the methods. R code is provided for convenience as well. The five-region methods are recommended for routine use during the analysis of epidemiologic data.

Adding a nonpainful end to reduce pain recollection of Pap smear screening: a randomized controlled trial

“…We used the 5-region hypothesis test based on odds ratios (ORs) to test the significant indistinguishability of real-time pain between the 2 groups for confirming the success of randomization and blinding. 18,33 The effect sizes of 5-minute recalled pain on a 1 to 5 numeric scale and a 0 to 10 VAS between the 2 groups were calculated using Glass delta. Linear regression analyses were used to estimate the effect of the intervention on scores of 5-minute and 1-year recalled pain.…”

A randomized, double-blind water taste test to evaluate the equivalence of taste between tap water and filtered water in the Taipei metropolis

“…A newly proposed five-region hypothesis test 18 was conducted to test the OR. The five regions were defined as (a recognizable ability to distinguish between the water varieties; we also consider a more lenient criterion of for this category), (a negligible ability to distinguish between the water varieties; also a stricter criterion of for this category), (no ability to distinguish between the water varieties), (a weak tendency to guess incorrectly), and (a strong tendency to guess incorrectly).…”

“…The five regions were defined as (a recognizable ability to distinguish between the water varieties; we also consider a more lenient criterion of for this category), (a negligible ability to distinguish between the water varieties; also a stricter criterion of for this category), (no ability to distinguish between the water varieties), (a weak tendency to guess incorrectly), and (a strong tendency to guess incorrectly). The 95% five-region confidence interval 18 of the OR was also calculated. To conclude 18 that the participants have no recognizable ability to distinguish between the two water varieties ( ), at least 207 participants were required to be recruited to achieve a power of 80% at a significance level of 0.05.…”

“…The 95% five-region confidence interval 18 of the OR was also calculated. To conclude 18 that the participants have no recognizable ability to distinguish between the two water varieties ( ), at least 207 participants were required to be recruited to achieve a power of 80% at a significance level of 0.05. Furthermore, we generated 10,000 bootstrap samples and calculated a 95% bootstrap confidence interval 19 of the AUC.…”