Alternatives to the statistical mass confusion of testing for no-effect

Abstract

Statisticians and researchers have argued about the merits of effect size estimation relative to hypothesis testing for decades. Cell biology has largely avoided this debate and is now in a quantitation crisis. In experimental cell biology, statistical analysis has grown to mean testing the null hypothesis that there was no experimental effect. This weak form of hypothesis testing neglects effect size, is universally misinterpreted, and is disastrously prone to error when combined with high-throughput cell biology. The first part of the solution proposed here is to limit statistical hypothesis testing to the small subset of experiments where a biologically meaningful null hypotheses can be defined prior to the experiment. The second part of the solution is to make confidence intervals the default statistic in cell biology.