Speaker: Prof. Davood Tofighi

Enhancing Causal Inference in Health and Psychological Research with Advanced Sensitivity Analysis

Dr. Tofighi will be presenting his research to the Department of Mathematics and Statistics in preparation for his Transfer from the Department of Psychology.

Abstract
Causal mediation analysis is a popular statistical technique employed to explore the causal relationships between variables in educational and psychological research. It enables the investigation of how an antecedent variable influences an outcome through an intermediate variable. However, the effectiveness of mediation analysis relies on the challenging assumption of the absence of unmeasured confounders, a condition often difficult to verify in real educational settings. In this presentation, I introduce an innovative sensitivity analysis approach specifically designed for nonrandomized latent growth curve mediation models (LGCMMs), which are frequently encountered in health and psychological research. This novel technique, named “Extended Correlated Augmented Mediation Sensitivity Analysis (ECAMSA),” extends a previous sensitivity analysis method to accommodate the distinctive characteristics of nonrandomized LGCMMs in educational and psychological studies. Through an empirical example, I demonstrate how ECAMSA can be applied to assess the robustness of mediation analysis results when faced with potential violations of the no-omitted confounder assumption. I argue that ECAMSA represents a valuable tool for enhancing the precision of causal inference in the field of educational research.

Biography

Davood Tofighi is an Associate Professor of Psychology at the University of New Mexico. His research focusses on developing and evaluating statistical methodologies that enhance the robustness of statistical inferences in various analytical domains. He is particularly interested in causal inference, application, and development of direct acyclic graph (DAG) in causal mediation analysis, statistical and Monte Carlo sensitivity analysis, generalized linear mixed-effects model, multivariate statistics, categorical data analysis, and the judicious application of machine learning techniques to health data.

Dr. Tofighi’s substantive research focuses on amplifying the replicability of statistical methods in the context of educational, health and prevention science. He is also interested in optimizing drug delivery formulations, a critical area of study with far-reaching implications for cancer and COVID-19 research. As an Associate Editor for the Quantitative Psychology and Measurement specialty section of Frontiers in Psychology, Dr. Tofighi plays an active role in the academic community, ensuring the dissemination of rigorous research to a broader audience. His contributions have garnered notable recognition, with a total of 7,659 citations and an H-Index of 16 according to Google Scholar. Dr. Tofighi acknowledges the support of funding institutions such as NASA, NSF, and NIH, which have been instrumental in enabling his research pursuits. His academic journey reflects his multidisciplinary approach, with a Ph.D. in Quantitative Psychology from Arizona State University, complemented by an M.A. in Educational Psychology. Additionally, he has earned M.Sc. and B.Sc. degrees in Industrial and Systems Engineering.