Teaching confidence intervals: Problems and potential solutions


Authors: 
Fiona Fidler & Geoff Cumming
Pages: 
online
Year: 
2005
Publisher: 
International Statistical Institute,
URL: 
http://www.stat.auckland.ac.nz/~iase/publications/13/Fidler-Cumming.pdf
Abstract: 

There are benefits of teaching inference via confidence intervals (CIs), rather than null<br>hypothesis significance testing (NHST). However, CIs are not without misconceptions.<br>First, we provide empirical evidence that CI presentations of data can help alleviate some<br>typical misinterpretations of results, leading to more accurate conclusions and more justified<br>decisions. Second, we demonstrate that CIs are also prone to particular types of<br>misconceptions. Finally, we present interactive figures and simulations that, when used with<br>guidelines for CI interpretation, should lead to more insightful interpretations of research<br>results and fewer misconceptions.