The lack of reproducibility in research: How statistics can endorse results

Scientific research is validated by reproduction of the results, but efforts to reproduce spurious claims drain resources. We focus on one cause of such failure: false positive statistical test results caused by random variability. Classical statistical methods rely on p-values to measure the eviden...

Full description

Saved in:
Bibliographic Details
Main Authors: Scott Goddard, Valen Johnson
Format: Article
Language:Catalan
Published: Universitat de València 2015-04-01
Series:Mètode Science Studies Journal: Annual Review
Subjects:
Online Access:https://turia.uv.es/index.php/Metode/article/view/3913
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1846164078621360128
author Scott Goddard
Valen Johnson
author_facet Scott Goddard
Valen Johnson
author_sort Scott Goddard
collection DOAJ
description Scientific research is validated by reproduction of the results, but efforts to reproduce spurious claims drain resources. We focus on one cause of such failure: false positive statistical test results caused by random variability. Classical statistical methods rely on p-values to measure the evidence against null hypotheses, but Bayesian hypothesis testing produces more easily understood results, provided one can specify prior distributions under the alternative hypothesis. We describe new tests, UMPBTs, which are Bayesian tests that provide default specification of alternative priors, and show that these tests also maximize statistical power. 
format Article
id doaj-art-c50a514203d347ab83e480fdb7f5ebb6
institution Kabale University
issn 2174-3487
2174-9221
language Catalan
publishDate 2015-04-01
publisher Universitat de València
record_format Article
series Mètode Science Studies Journal: Annual Review
spelling doaj-art-c50a514203d347ab83e480fdb7f5ebb62024-11-18T16:02:56ZcatUniversitat de ValènciaMètode Science Studies Journal: Annual Review2174-34872174-92212015-04-01510.7203/metode.0.3913The lack of reproducibility in research: How statistics can endorse resultsScott Goddard0Valen Johnson1<p>Texas A&amp;M Universitiy (USA).</p><p>Texas A&amp;M Universitiy (USA).</p>Scientific research is validated by reproduction of the results, but efforts to reproduce spurious claims drain resources. We focus on one cause of such failure: false positive statistical test results caused by random variability. Classical statistical methods rely on p-values to measure the evidence against null hypotheses, but Bayesian hypothesis testing produces more easily understood results, provided one can specify prior distributions under the alternative hypothesis. We describe new tests, UMPBTs, which are Bayesian tests that provide default specification of alternative priors, and show that these tests also maximize statistical power.  https://turia.uv.es/index.php/Metode/article/view/3913statistical evidencehypothesis testBayesian analysisuniformly most powerful Bayesian tests
spellingShingle Scott Goddard
Valen Johnson
The lack of reproducibility in research: How statistics can endorse results
Mètode Science Studies Journal: Annual Review
statistical evidence
hypothesis test
Bayesian analysis
uniformly most powerful Bayesian tests
title The lack of reproducibility in research: How statistics can endorse results
title_full The lack of reproducibility in research: How statistics can endorse results
title_fullStr The lack of reproducibility in research: How statistics can endorse results
title_full_unstemmed The lack of reproducibility in research: How statistics can endorse results
title_short The lack of reproducibility in research: How statistics can endorse results
title_sort lack of reproducibility in research how statistics can endorse results
topic statistical evidence
hypothesis test
Bayesian analysis
uniformly most powerful Bayesian tests
url https://turia.uv.es/index.php/Metode/article/view/3913
work_keys_str_mv AT scottgoddard thelackofreproducibilityinresearchhowstatisticscanendorseresults
AT valenjohnson thelackofreproducibilityinresearchhowstatisticscanendorseresults
AT scottgoddard lackofreproducibilityinresearchhowstatisticscanendorseresults
AT valenjohnson lackofreproducibilityinresearchhowstatisticscanendorseresults