Research Papers

Choosing a Test to Predict Driving Abilities: Do Not Rely Only on Statistical Associations

Filename cmrsc19_35.pdf
Filesize 64 KB
Version 1
Date added June 7, 2009
Downloaded 2 times/fois
Category 2009–CMRSC-XIX–Saskatoon
Tags Session 2B
Author/Auteur Michel Bédard, Bruce Weaver, Peteris Darzins, Michelle M. Porter

Background

P-values (e.g., “p < .05”) are often interpreted as a sign of important research findings. However, p-values are affected by a number of parameters such as effect size and sample size, and so cannot be used to assess clinical or practical importance. It is not unusual to see research findings where an apparently important difference between groups is evident yet fails to be statistically significant. This results in the claim that there is no difference between groups. On the other hand, differences between groups can be minuscule and unimportant, yet heralded as important, simply because the p-value was less than .05. Such situations often arise from the use of very large sample sizes, and while the difference is not due to the play of chance, the real life implications are non-existent.

Michel Bédard, Bruce Weaver, Peteris Darzins, Michelle M. Porter