Effect size and precision in psychological research: The use of confidence intervals

Abstract

Interpretations of data are suggested, in which the use of the null-hypothesis significance testing is avoided. The interpretations are based solely on considering effect size and precision of measurement. A “good” precision is quantified as the confidence interval that corresponds to an effect size of no practical importance. Methods for evaluating post hoc precision as well as planning the sample size for a desired precision are presented for the cases of difference between independent means, a single mean, difference between paired means, single proportion and product-moment correlation coefficient. The relation between analysis of precision and power analysis is discussed.