CMStatistics 2018: Start Registration
View Submission - CMStatistics
B0381
Title: Towards a new R package for fitting Poisson linear excess relative risk models Authors:  Manuel Higueras - Basque Center for Applied Mathematics (Spain) [presenting]
Abstract: The excess relative risk (ERR) represents the elevated rate of disease (e.g., cancer) per unit of exposure (e.g., ionizing radiation). Relative risk models are typically applied in radiation epidemiology follow-up studies, for instance those which analyze the risk of leukemia or brain tumor in pediatric patients who have been examined with computed tomography scans. It is proposed the fitting of general Poisson linear relative risk models in R as an alternative of EPICURE's AMFIT module, which is the gold standard in radiation epidemiology practice. This fitting is performed by means of the maxLik R package and an efficient implementation of the gradient of the log-likelihood with respect the parameters. These Poisson linear ERR models require data in person-years format with time-depending and cumulative variables, and they can be built in R by means of the pyears function at the survival package, as an alternative of EPICURE's DATAB module. The aim is to join these two tools, person-years table builder and Poisson linear ERR model fitter, in an R package to give a free license and open source alternative to EPICURE, a very specialized proprietary software.