In this talk, students and participants will gain insight into vital medical research and observe how traditional statistical methods can be modified to better fit the given data.
Monday, December 3rd
Speaker: Ramon Durazo-Arvizu (LUC – Stritch School of Medicine)
Title: Modeling the Association between 25(OH)D and Mortality in a Representative US Population Sample
Lecture: 4:30 p.m., Cuneo Hall 312
Meet the Speaker: 4:00 p.m., Cuneo Hall 312
w/ tea & cookies
Regression techniques have been introduced to assist researchers to find relationships between key variables, and these methods have been especially useful in epidemiological studies such as the NHANES (National Health and Nutrition Examination Survey). Traditional regression approaches relied upon the Normality distribution fitting the response variable, and these have been extended to other distributions, such as for the Binomial Logistic Model. A novel approach is to use piecewise splines, or segmented polynomial curves pieced together to better fit the underlying data. In this talk, we focus on data from the Third National Health and Nutrition Examination Survey (NHANES III) with an eye to characterizing the relationship between serum 25-hydroxyvitamin D [25(OH)D] and all-cause mortality using statistical modelling approaches. Specifically, using multivariate logistic regression, two possible analytic strategies were compared: (1) transformation of 25[OH]D to normality and (2) restricted cubic splines – after adjusting for key covariates. For these data, the cubic splines approach resulted in lower estimates and narrow CI. As the Normal transformation approach assumes a symmetrical model (which was not observed in these data), the cubic splines approach may be more appropriate.
About the Speaker: Ramon Durazo-Arvizu is a professor and head of Section of Biostatistics at the Preventive Medicine & Epidemiology Department at the Loyola’s Stritch School of Medicine. His duties include coordinating and training staff of statisticians and computer scientists; consultation with medical and basic-science investigators on design, implementation, and analysis of research studies; and conducting computer-based and analytic research. Dr Durazo-Arvizu received his B.S. in Mathematics from Universidad Autonoma de Puebla, Puebla, Mexico; M.S. in Statistics from University of Texas, El Paso; and Ph.D. in Applied Mathematics from University of Arizona, Tucson. He is currently the principal and co-principal investigator in several projects funded by the National Institute of Health.
More Details: http://www.luc.edu/math/ucms/