Statistical methods for population research have advanced a long way from the two sample t test (though sometimes that familiar test is still exactly the tool required to answer the research question!). In this talk I will show how much of new methodological research in biostatistics revolves around two important concepts - understanding variation, and the linear model. I will present a unified view of such concepts as multilevel models for clustered data, random effects models for longitudinal data, autocorrelated models for spatial data, and multiple imputation for missing data. Examples will be drawn from population research that highlight the different ways these models conceptualise variation, as well as how they all unify under the concept of the linear model.

Alice Richardson studied at Victoria University of Wellington, New Zealand, then at the Australian National University in the Department of Applied Statistics. Her PhD was on the statistical properties of robust methods of estimation for multilevel linear models. Alice then accrued twenty years of experience in teaching undergraduate statistics at the University of Canberra. During that time she also collaborated on quantitative research projects in a diverse range of disciplines including Nutrition, Sports Science, Linguistics and Management. In 2016 she took up a position as biostatistician at the National Centre for Epidemiology & Population Health. Her research interests are in linear models and robust statistics; statistical properties of data mining methods; applications of statistical methods to large data sets especially in population health and the biomedical sciences; and innovation in statistics education.