An Exciting Alternative to Fisher's Exact Test for Two Proportions

Abstract

When α definite decision must be made as to which of two or more populations is best, a selection procedure seems more appropriate than a significance test. For the common problem of selecting the better one of two populations with regard to some proportion, a table or curve gives the sample size required for 90% confidence of a correct selection. Comparison with Fisher's Exact Test, the best significance test for two sample proportions, shows that the selection procedure is simpler and requires a much smaller sample size.

• Binomial Distribution
• Fisher's Exact Test
• Ranking and Selection

Additional information

Notes on contributors

Wendell E. Carr

Mr. Carr is an Advisory Statistician with International Business Machines. He is a Senior Member of ASQC.