Hypothesis testing - Proportions

We have described elementary methods for testing hypotheses about proportions for 1 or 2 populations. These methods are used for dichotomous data and include a z test based on the normal distribution to compare 1 sample proportion with a postulated or reference value, as well as χ2 tests to compare proportions in 2 independent or paired samples. With small samples or proportions close to 0 or 1, exact tests should be used instead of these large-sample approximate procedures. Methods to account for concomitant or confounding variables will be addressed later in the series. © 2006 American Heart Association, Inc.