A1589
Title: Assessing methods for detecting outliers in meta-analysis
Authors: Samuel Manda - University of Pretoria (South Africa)
Memory Makuta - University of Malawi (Malawi) [presenting]
Abstract: Outlier data sets in meta-analyses can have a substantial negative impact on the validity of empirical findings and the strength of the conclusions. Detecting outliers and the ways to handle them are crucial in ensuring the reliability of meta-analytic findings. The random effects variance shift model is discussed in comparison to the more common ways of outlier detection in meta-analyses. These methods will be analytically compared for their ability to identify outliers and offer insights into their strengths and limitations. The theoretical evaluation of methods is supplemented by extensive simulation studies. Potential outlier data sets are then identified in a meta-analysis of the prevalence of unhealthy food consumption among children aged 6 to 23 months using several Demographic and Health Survey datasets from 2010 to 2022 across sub-Saharan African countries
