Abstract
Estimation, based on effect sizes (ESs) and confidence intervals (CIs), is much better than null hypothesis significance testing (NHST). I refer to estimation and meta‐analysis—which is the extension of estimation to multiple studies—as the new statistics. The techniques themselves are not new, but using them would for many psychologists be new, and a highly beneficial advance. I describe a six‐step strategy for estimation, which starts with the statement of research questions in terms of “how large?” questions, rather than the dichotomous hypotheses of NHST. I outline how to use estimation to analyse the two‐independent‐groups and paired designs, and randomised control trials. I discuss the ES measures Cohen's d and correlation, r, in relation to estimation. I describe how to interpret results published using NHST by visualising the corresponding CIs, and give guidance for adopting the new statistics, even in a world that often still expects NHST. I emphasise the value of figures that display CIs, and describe freely available software running under Microsoft Excel that assists calculation of CIs and preparation of figures.
This research was supported by the Australian Research Council. I thank Jacenta Abbott and Laura Anderson for valuable comments on a draft.
This research was supported by the Australian Research Council. I thank Jacenta Abbott and Laura Anderson for valuable comments on a draft.
Notes
This research was supported by the Australian Research Council. I thank Jacenta Abbott and Laura Anderson for valuable comments on a draft.