Why do ecologists aim to get positive results? Once again, negative results are necessary for better knowledge accumulation

Abstract

Hypothesis testing is commonly used in ecology and conservation biology as a tool to test statistical–population parameter properties against null hypotheses. This tool was first invented by lab biologists and statisticians to deal with experimental data for which the magnitude of biologically relevant effects was known beforehand. The latter often makes the use of this tool inadequate in ecology because we field ecologists usually deal with observational data and seldom know the magnitude of biologically relevant effects. This precludes us from using hypothesis testing in the correct way, which is posing informed null hypotheses and making use of a priori power tests to calculate necessary sample sizes, and it forces us to use null hypotheses of equality to zero effects which are of little use for field ecologists because we know beforehand that zero effects do not exist in nature. This is why only ‘positive’ (statistically significant) results are sought by ecologists, because negative results always derive from a lack of power to detect small (usually biologically irrelevant) effects. Despite this, ‘negative’ results should be published, as they are important within the context of meta–analysis (which accounts for uncertainty when weighting individual studies by sample size) to allow proper decision–making. The use of multiple hypothesis testing and Bayesian statistics puts an end to this black or white dichotomy and moves us towards a more realistic continuum of grey tones.