Error Rates, Decisive Outcomes and Publication Bias with Several Inferential Methods.

BACKGROUND

Statistical methods for inferring the true magnitude of an effect from a sample should have acceptable error rates when the true effect is trivial (type I rates) or substantial (type II rates).

OBJECTIVE

The objective of this study was to quantify the error rates, rates of decisive (publishable) outcomes and publication bias of five inferential methods commonly used in sports medicine and science. The methods were conventional null-hypothesis significance testing [NHST] (significant and non-significant imply substantial and trivial true effects, respectively); conservative NHST (the observed magnitude is interpreted as the true magnitude only for significant effects); non-clinical magnitude-based inference [MBI] (the true magnitude is interpreted as the magnitude range of the 90 % confidence interval only for intervals not spanning substantial values of the opposite sign); clinical MBI (a possibly beneficial effect is recommended for implementation only if it is most unlikely to be harmful); and odds-ratio clinical MBI (implementation is also recommended when the odds of benefit outweigh the odds of harm, with an odds ratio >66).

METHODS

Simulation was used to quantify standardized mean effects in 500,000 randomized, controlled trials each for true standardized magnitudes ranging from null through marginally moderate with three sample sizes: suboptimal (10 + 10), optimal for MBI (50 + 50) and optimal for NHST (144 + 144).

RESULTS

Type I rates for non-clinical MBI were always lower than for NHST. When type I rates for clinical MBI were higher, most errors were debatable, given the probabilistic qualification of those inferences (unlikely or possibly beneficial). NHST often had unacceptable rates for either type II errors or decisive outcomes, and it had substantial publication bias with the smallest sample size, whereas MBI had no such problems.

CONCLUSION

MBI is a trustworthy, nuanced alternative to NHST, which it outperforms in terms of the sample size, error rates, decision rates and publication bias.