Use this URL to cite or link to this record in EThOS:
Title: A review of multiple hypothesis testing in relation to the use of lateral cephalometric variables as the outcome measure in orthodontic research
Author: Pua, Shih Chia
ISNI:       0000 0004 7659 0932
Awarding Body: University of Liverpool
Current Institution: University of Liverpool
Date of Award: 2018
Availability of Full Text:
Access from EThOS:
Access from Institution:
Aim: To examine the extent of the multiple hypothesis testing and its correction in orthodontic research in relation to the use of lateral cephalometric variables as the outcome measure. Study design: A retrospective, observational study looking at a sample of published orthodontic articles (n=1688) over a two-year period from 1st January 2014 to 31st December 2015. Data sources: Four major electronic databases namely PubMed, Ovid Medline, Scopus and EBSCO Dentistry & Oral Sciences Source were electronically searched using Medical Subject Heading (MeSH) terms. Additionally, all issues of American Journal of Orthodontics and Dentofacial Orthopedics (AJODO), The Angle Orthodontist (AO), European Journal of Orthodontics (EJO) and Journal of Orthodontics (JO) were also hand-searched systematically. Both searches were carried out independently by first author (SCP). Review methods: Eligible articles were identified and reviewed independently by first author (SCP) to determine whether the articles tested greater than five hypotheses in at least one family of inferences with respect to the predetermined criteria. For articles meeting the criterion for multiple testing, type I error rates were calculated. Additionally, a statistical correction experiment using Bonferroni's method was applied to the reported results of the included studies. Additional information was collected on: study type (prospective/retrospective), journal classification (main/non-main orthodontic journal were classified based on 2015 SCImago Journal and Country Rank), region of authorship (Americas, Europe and Asia/others), number of researcher in the publication (1-4, 5-7 and 8 or more) and involvement of a statistician to examine whether these factors were associated with multiple testing correction. Results: Of the 139 studies associated with multiple testing, there was approximately 3 families of tests (per article) with an average of 20 hypothesis tests (range 5-47) using lateral cephalometric variables as the outcome measure (per family of tests). Only 40 publications (29%) considered the effect of multiple testing that these studies in some way have corrected or accounted for multiple testing. Within the studies that have not accounted for multiple hypothesis testing, there was a mean 58% chance of committing a type I error and, on average, 13% of the significant results were likely to be false positives. After the application of the Bonferroni's method in the correction experiment, only 47% of the significant results reported within the articles that remained significant. Studies published in the main orthodontic journals (AJODO, EJO, JO and KJO) were more likely to account for multiple testing (p=0.002). Handsearching was superior than electronic searching with 10% of papers (n=5) which were missed from electronic searching. Conclusions: Multiple testing is common in the orthodontic research especially in relation to the use of multiple cephalometric variables as the outcome measure. This study demonstrates that the risk of false positive findings is considerably high and only a minority of the articles that have in some way corrected or accounted for multiple testing. Therefore, this multiplicity issue in relation to the use of multiple cephalometric variables in a cephalometric study deserves a closer attention from researchers, reviewers and readers.
Supervisor: Burnside, Girvan ; Flannigan, Norah Sponsor: Not available
Qualification Name: Thesis (D.D.Sc.) Qualification Level: Doctoral