On 4 December 2014, the Department of Statistics at the University of Warwick hosted an RSS West Midlands local group meeting. The speakers were Jennifer Rogers from the London School of Hygiene and Tropical Medicine, and Glenn Woodcock, founder and director of the research company SPARX.
Jennifer’s talk, 'Insights into the analysis of recurrent events and future directions', began by illustrating existing survival analysis models by applying them to the chronic heart failure studies EMPHASIS-HF and CHARM-Preserved. She presented composite endpoint framework’s difficulties, and contrasted the treatment effect under such composite endpoint model, and Andersen-Gill model, and a negative binomial mixture model which considers a gamma distributed frailty term. Although Cox proportional hazards is a popular model in medical survival studies, the proportionality assumption is often violated for these kinds of data. The negative binomial mixture model provides an alternative, since it belongs to the family of accelerated life models.
Jennifer described her current and future work on diverse aspects of the modelling. One of such projects provides a sample size formula in order to minimize the number of patients needed in a study of recurrent events. The sample sizes derived from the negative binomial mixture model appear to be consistently lower than the ones provided by Cox proportional hazard’s model. Finally, Jennifer presented a semiparametric joint frailty model which considers a patient-specific frailty term, and measures the impact of the frailty itself.
In the second part of the group meeting, Glenn began his talk, 'Taming feedback with technology', by introducing the inherent difficulties of measuring and studying the learning process of students while they work. Since different types of feedback are received very quickly, a system is designed in order to store the information and analyse the student’s learning skills simultaneously.
The data is collected when the students are given a set of problems with model solutions, but is complemented by a set of available educational games that the students can access along the examinations. The purpose of the obtained feedback was: to create a repetition for the students to trigger a reflex to certain types of questions; to provide reinforcement and encouragement when solving problems; to build the student’s confidence; and correct their behaviour towards some types of exercises.
Glenn concluded his talk by explaining that the immediate interests of the SPARX system developers lies in improving the teaching and learning methods for the students, by means of addressing a list of important questions. The system has been implemented and used in a certain number of classrooms in some schools, and the students and professors have been randomised as much as possible, but as Glenn mentioned, further development of questions and models are being explored in order to measure the learning aspects of interest.