File(s) under permanent embargo
Review of null hypothesis significance testing in the ophthalmic literature: are most 'significant' P values false positives?
journal contribution
posted on 2023-05-18, 12:17 authored by Sanfilippo, PG, Casson, RJ, Yazar, S, David MackeyDavid Mackey, Alexander HewittAlexander HewittP values associated with null hypothesis significance testing (NHST) are almost universal in the ophthalmic literature. A P value < 0.05 is traditionally considered 'significant'. This concept may deflect further thought about the veracity of the results. P values influence the publishability of the data and have flow-on effects for funding success and the direction of future research. Despite their importance, the problems inherent in P values have been recognized since their inception, and in more recent years have been increasingly highlighted in some scientific fields. In this review, we aim to bring the problems associated with P values and NHST to the attention of the ophthalmic research community. We do not offer a universal solution to the problem of determining the veracity of a scientific claim; however, we demonstrate the need for caution in interpreting 'significant' P values by performing a Bayesian re-analysis of t-tests in the ophthalmic literature.
History
Publication title
Clinical and Experimental OphthalmologyVolume
44Pagination
52-61ISSN
1442-6404Department/School
Tasmanian School of MedicinePublisher
Blackwell Publishing AsiaPlace of publication
54 University St, P O Box 378, Carlton, Australia, Victoria, 3053Rights statement
Copyright 2015 Royal Australian and New Zealand College of OphthalmologistsRepository Status
- Restricted