Research Article

Environmental Science and Pollution Research

, Volume 20, Issue 10, pp 7341-7347

Misuse of null hypothesis significance testing: would estimation of positive and negative predictive values improve certainty of chemical risk assessment?

  • Mirco BundschuhAffiliated withInstitute for Environmental Sciences, University of Koblenz-LandauDepartment of Aquatic Sciences and Assessment, Swedish University of Agricultural Sciences Email author 
  • , Michael C. NewmanAffiliated withVirginia Institute of Marine Science, College of William and Mary
  • , Jochen P. ZubrodAffiliated withInstitute for Environmental Sciences, University of Koblenz-Landau
  • , Frank SeitzAffiliated withInstitute for Environmental Sciences, University of Koblenz-Landau
  • , Ricki R. RosenfeldtAffiliated withInstitute for Environmental Sciences, University of Koblenz-Landau
  • , Ralf SchulzAffiliated withInstitute for Environmental Sciences, University of Koblenz-Landau

Rent the article at a discount

Rent now

* Final gross prices may vary according to local VAT.

Get Access

Abstract

Although generally misunderstood, the p value is the probability of the test results or more extreme results given H0 is true: it is not the probability of H0 being true given the results. To obtain directly useful insight about H0, the positive predictive value (PPV) and the negative predictive value (NPV) may be useful extensions of null hypothesis significance testing (NHST). They provide information about the probability of statistically significant and non-significant test outcomes being true based on an a priori defined biologically meaningful effect size. The present study explores the utility of PPV and NPV in an ecotoxicological context by using the frequently applied Daphnia magna reproduction test (OECD guideline 211) and the chemical stressor lindane as a model system. The results indicate that especially the NPV deviates meaningfully between a test design strictly following the guideline and an experimental procedure controlling for α and β at the level of 0.05. Consequently, PPV and NPV may be useful supplements to NHST that inform the researcher about the level of confidence warranted by both statistically significant and non-significant test results. This approach also reinforces the value of considering α, β, and a biologically meaningful effect size a priori.

Keywords

Sample size Bayesian Power analysis Effect size Type I error rate Type II error rate