Implementation science and research methods seminar: Multilevel modelling – Stellamay Gwini

The latest CaHRU Implementation Science and Research Methods seminar was given on 1 September 2017 by guest lecturer and statistician, StellaMay Gwini of Monash University, Melbourne, Australia. StellaMay is a Research Fellow (Biostatistician) with the Department of Epidemiology and Preventive Medicine at Monash . She has a background in applied medical research having commenced her career with the Lincolnshire National Health Service gwiniin the UK with Professor Siriwardena where she worked primarily on a case-control and self-controlled case series study funded by the NIHR Research for Patient Benefit programme exploring the relationship between flu vaccination and protection against heart attack (acute myocardial infarction) using GPRD data. The studies were published in the Canadian Medical Association Journal and Vaccine. Her main interests are in appropriate use of both statistical methods and research techniques for health service/technology use, epidemiology and public health research.

StellaMay spoke on the topic of multilevel modelling. She described how multilevel modelling, also called hierarchical or nested modelling can be used where subjects are clustered into groups giving rise to the possibility that measured outcomes might be correlated. Usual regression techniques assume that data points are independent hence these methods are not appropriate for hierarchical data or repeated measurements (data collected multiple times for the same subject) where independence cannot be assumed.

Hence multilevel modelling allows for correct inferences to be drawn from data, without making the assumption of independence. Stellamay concluded by describing a recent study which she has conducted on efficacy of raloxifene therapy for schizophrenia treatment, the presentation  highlight the usefulness of multilevel modelling in medical research. Her talk was informative and very well received.

Leave a Reply

Your email address will not be published. Required fields are marked *