Date/Time
Date(s) - 02/04/2024
3:30 pm - 4:30 pm

Date/Time:  April 2, 2024 (3:30PM – 4:30PM)

Location:  KTH B105

Speaker: Zelalem Negeri (Assistant Professor, Department of Statistics and Actuarial Science, University of Waterloo)

Title: A bivariate finite mixture random effects model for identifying and accommodating outliers in diagnostic test accuracy meta-analyses

Abstract: Outlying studies are prevalent in meta-analyses of diagnostic test accuracy studies and may lead to misleading inferences and decision-making unless their negative effect is appropriately dealt with. Statistical methods for detecting and downweighting the effect of such studies have recently gained the attention of many researchers. However, these methods dichotomize each study in the meta-analysis as outlying or nonoutlying and focus on examining the effect of outlying studies on the summary sensitivity and specificity only. We developed and evaluated robust and flexible random effects bivariate finite mixture models for meta-analyzing diagnostic test accuracy studies. The proposed model accounts for both the within- and across-study heterogeneity in diagnostic test results, generates the probability that each study in a meta-analysis is outlying instead of dichotomizing the status of the studies, and allows assessing the impact of outlying studies on the pooled sensitivity, pooled specificity, and between-study heterogeneity. Our simulation study and real-life data examples demonstrated that the proposed model was robust to the existence of outlying studies, produced precise point and interval estimates of the pooled sensitivity and specificity, and yielded similar results to the standard model when there were no outliers. Researchers and practitioners may use our proposed model as a stand-alone model or an alternative to the standard model to conduct sensitivity analysis when outlying studies are present in a meta-analysis.

Bio:  Dr. Zelalem Negeri is an Assistant Professor at the Department of Statistics and Actuarial Science at the University of Waterloo. He earned Ph.D. (2019) and MSc (2015) degrees in Statistics from McMaster University under the supervision of Professor Joseph Beyene before completing a two-year postdoctoral fellowship at McGill University in 2021. He has also received an MSc (2014) and a BSc (2011) degree in statistics from Addis Ababa University in Ethiopia. Dr. Negeri’s research interest focuses on developing and validating statistical methods for applications in public health research, emphasizing both aggregate data and individual participant data meta-analyses of diagnostic and screening test accuracy studies, utilizing computational statistics methods such as parametric and non-parametric bootstrap approaches and deterministic and Monte Carlo expectation-maximization algorithms. Dr. Negeri teaches statistics courses and supervises students at the University of Waterloo.