Statistical and other methodological issues in the production of reliable experimental endodontic research and why they matter
| Autor: | Donald M. Brunette |
|---|---|
| Rok vydání: | 2016 |
| Předmět: |
0301 basic medicine
Scope (project management) business.industry False positives and false negatives Replicate computer.software_genre Data science Bayesian statistics 03 medical and health sciences 030104 developmental biology 0302 clinical medicine Scientific method Medicine Data mining business Null hypothesis computer Publication 030217 neurology & neurosurgery Statistical hypothesis testing |
| Zdroj: | Endodontic Topics. 34:8-29 |
| ISSN: | 1601-1538 |
| Popis: | It has long been assumed that science is self-correcting and that lack of reproducibility or even fraud will be uncovered in subsequent attempts to replicate observations or experiments. Recent studies have cast doubt on that assumption, which is a central tenet of scientific method, even for research that is published in high-impact journals and uses state-of-the-art methodology. Ironically, endodontic research, like most biomedical research, is moving toward highly sophisticated procedures such as gene arrays and proteomics, which are often plagued by false positives and false negatives. This “reproducibility crisis” is thought to arise from many sources including poorly described methodology, inappropriate statistical analysis, the positive bias for article acceptance by journals, the lack of incentives to repeat experiments, and the hypercompetitive research environment that pressures investigators to publish quickly in impact journals. This article outlines the scope of the reproducibility crisis and some of the steps that may be used to ameliorate it. Attention is focused on common statistical errors and pitfalls, the use of confidence intervals in place of the standard Null Hypothesis Statistical Testing (NHST), and the promise of Bayesian statistics in endodontic research. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Plný text