Consider the gastroenteritis and beef consumption data presented in Table 6.8. The relative risk is 5.7, which most epidemiologists would deem a “strong” association between exposure and disease. In addition, the p-value is exceedingly small, less than 0.001, and far less than the commonly used cutoff of 0.05. So the investigators rejected the null hypothesis (that beef was not associated with illness) and adopted the alternative hypothesis (that beef was indeed associated with illness). In this outbreak, the association between eating beef at the banquet and gastroenteritis was both strong (RR=5.7) and statistically significant (p < 0.001).
The chi-square test works well if the number of people in the study is greater than about 30. For smaller studies, a test called the Fisher Exact Test may be more appropriate. Because the Fisher Exact Test is tedious to calculate, let Epi Info or another computer program perform the calculations for you.