Counting by weighing and its effect on comparing population proportions

Counting by weighing is a rapid procedure for obtaining N items (e.g., seeds) based on weight rather than exact counting. We show how to modify the test statistic used for comparing m population proportions (e.g., seed germination rates) when independent samples are taken using this approximate counting procedure. The modified test will have approximate size [alpha] as advertised under the null hypothesis that there is no difference among the m proportions. A simulation study is conducted to determine the small to moderate sample size Type I error rates of the classical and modified tests.