To the Editor.
—Spach et al1 reported a case-control study of 78 young men and concluded that lack of circumcision increases the risk of urinary infection. Their "Statistical Analysis" section of the report consisted of the single sentence, "Comparison between bacteriuric and nonbacteriuric groups were made by χ2 analysis."This is a rather nonspecific statement given that there are a multitude of tests that are useful for 2×2 contingency tables that have an asymptotic χ2 distribution. These include the Pearson χ2 test and the likelihood ratio G2 statistic. It is well known that these asymptotic tests can give only approximate results when there is a small total sample size or small numbers in individual cells. One rule often stated is that expected cell sizes should exceed four or five. However, Fienberg states that "obvious exceptions to the rule of 'average cell sizes of 4 or