Unit:
Κατεύθυνση ΒιοστατιστικήLibrary of the School of Health Sciences
Author:
Balasopoulos Ioannis
Supervisors info:
Ντζούφρας Ιωάννης, Καθηγητής, Τμήμα Στατιστικής, ΟΠΑ
Νικολακόπουλος Σταύρος, Επίκουρος Καθηγητής, Tμήμα Ψυχολογίας, Πανεπιστήμιο Ιωαννίνων
Σιάννης Φώτιος, Επίκουρος Καθηγητής, Τμήμα Μαθηματικών, ΕΚΠΑ
Original Title:
Meta-analysis of bayes factors
Translated title:
Meta-analysis of bayes factors
Summary:
Bayes Factors is a statistical measure used in Bayesian statistics for hypothesis
testing. They prove to be far superior to the p-values and test statistics at
making inference. That is because they can provide evidence both for and
against the null hypothesis, and they have the potential to be less dependent on
sample size, as opposed to their frequentist counterpart. However, the
implementation of methods for combining Bayes Factors from multiple studies
has been limited in literature. In a recent study by Nikolakopoulos and
Ntzoufras (2021), they reviewed existing methods for calculating Bayes factors
in a single study and suggested methods for combining Bayes factors from
multiple studies into a single estimate, followed by simulation in order to
examine their performance. However, we need to take into consideration that
these methods were applied on a fixed-effect model while in reality in a metaanalysis, the random effects model is considered more suitable in describing
the data. This is because different studies provide realizations of effects that
come from a common a distribution (exchangeability trait), rather than a fixed
quantity, and are characterized by heterogeneity (Higgins et al., 2009). In this
study we tried to assess the performance of these methods under a different
simulation scenario, by adding some variability to the data in the form of none
fixed variance across the studies. Similarly to Nikolakopoulos and Ntzoufras
(2021) we conclude that regardless of the sample variance, the sample size and
statistic of the individual studies (likelihood, t-test, likelihood ratio or even
Bayes Factor) are enough to adequately obtain an overall Bayes Factor from a
number of studies. This can be attributed to the fact the difference in sample
variance may be offset by the different sample size of each study.
Main subject category:
Health Sciences
Keywords:
Bayes factors, Meta-analysis
File:
File access is restricted only to the intranet of UoA.
BALASOPOULOS FINAL (1).pdf
1 MB
File access is restricted only to the intranet of UoA.
