Summary
Contents
Subject index
This book is a simple introduction for nonstatisticians to power analysis and sample size determination. It clearly illustrates why sample sizes need to be sufficiently large, so that the experiment has good power properties and hence low type II error rates. The authors introduce a simple technique of statistical power analysis that allows researchers to compute approximate sample sizes and power for a wide variety of research designs. Because the same technique is used with only slight modifications for different statistical tests, researchers can easily compare the sample sizes required by different designs and tests to make cost-effective decisions in planning a study. These comparisons, emphasized throughout the book, demonstrate some important principles of design, measurement, and analysis that are not given sufficient emphasis in other texts.
Correlation Coefficients
Correlation Coefficients
In the last chapter, we considered tests of the equality of means. Frequently, however, research questions concern the association of variables rather than simply comparisons of means in one or more groups. For instance, rather than testing whether mean health status is higher among heavy coffee drinkers versus abstainers, we could ask a more general question: As coffee consumption increases, does health status decrease? Or vice versa? Formulating the question would carry the advantage that arbitrary cut points (heavy: over six cups a day; moderate: three to five cups a day; etc.) will not affect the conclusion. The most commonly used effect size to describe such an association is the Pearson product-moment correlation coefficient: ρP. While the two-sample t-test is arguably the ...
- Loading...