Dr. Kazeem Adepoju, University of Minnesota
224 Church St SE
Minneapolis,
MN
55455
Modeling Volatility in Length-Biased Survival Data: A Comparative Study of LB-GMM-GARCH and STAR-GARCH for Cancer Patient Outcomes
Abstract
This paper introduces a novel framework that integrates Length-biased Gaussian Mixture Models (LBGMM) with Generalized Autoregressive Conditional Heteroskedasticity (GARCH) models for clustering volatility regimes in cancer patient data. The proposed LBGMM-GARCH approach offers a flexible alternative to the Smooth Transition Autoregressive GARCH (STAR-GARCH) model by capturing diverse volatility patterns through probabilistic clustering of volatility regimes. By leveraging LBGMM-based clustering, the model avoids the deterministic structure of transition functions in STAR models, providing a more adaptive representation of health state fluctuations in cancer patients. Empirical analysis demonstrates that the LBGMM-GARCH model outperforms conventional regime-switching models in capturing volatility persistence, enhancing the understanding of dynamic health patterns and potential risks in cancer patient data.
Bio
Dr. Kazeem Adepoju is a Senior Lecturer and Assistant Director at the School of Statistics, University of Minnesota, USA, which he joined in 2018. He began his teaching career at the college level in 2012 and previously served as a Lecturer at the University of Ibadan, Nigeria. Dr. Adepoju earned a Professional Diploma in Statistics with Distinction, a Bachelor of Science in Statistics with First Class Honors, and both a Master of Science and Ph.D. in Statistics, all from the University of Ibadan. He completed a postdoctoral fellowship under the mentorship of Professor Galin Jones, Director of the School of Statistics at the University of Minnesota.
Dr. Adepoju’s research interests include Statistical Machine Learning, Mathematical Statistics, robust statistical models, statistics education, random effects estimation in repeated measures settings, longitudinal data modeling, and Machine Learning algorithm development. He has presented his work at numerous international conferences and authored approximately 25 research articles published in reputable scholarly journals.
He is actively involved in professional organizations such as the Nigerian Statistical Association, Nigerian Mathematical Society, American Mathematical Society, Mathematical Association of America, Bernoulli Society for Mathematical Statistics & Probability, and the American Statistical Association.
Over the years, Dr. Adepoju has received several prestigious research grants. Notable among these are funding for the 2013 ICOSDA conference at Central Michigan University, a grant from the Association of Commonwealth Universities in 2013 to present his Ph.D. work at the University of Oxford, United Kingdom, participation in the 2017 Joint Mathematics Meeting in Atlanta, Georgia, the 2018 Stochastic Processes and Applications conference at Brown University, and the 2019 Design and Analysis of Experiments workshop at the University of Tennessee, Knoxville. He continues to contribute significantly to research and academia through presentations and collaborations within and beyond the University of Minnesota.