|SPIDA 2012: The Program|
In its thirteenth season this year, ISR's Summer Program in Data Analysis focuses on linear models, beginning with “standard” regression, through generalized linear models and extending to mixed models, which incorporate two or more hierarchical levels of data or longitudinal data structures.
Linear models and their extension to generalized linear models (which unify linear models with other commonly employed statistical models, such as logistic and Poisson regression) are the workhorses of quantitative social research. Linear and generalized linear models additionally provide the basis for other, more advanced statistical techniques, including the mixed-effects models that are the focus of this year's SPIDA. The first part of SPIDA will introduce participants to the R statistical computing environment; review linear models, including their implementation in R, "diagnostic" methods for checking models fit to data, and the elliptical geometry of least-squares regression; and introduce the generalized linear models framework and its implementation in R. This part of the Program will be taught by Professor John Fox of McMaster University.
Linear models provide the basis for multilevel or mixed models, the topic of the second half of SPIDA 2012. Mixed models are useful for a wide range of data structures and research questions. They can be used for the analysis of hierarchical data, for example when students are nested in classes, which in turn are nested in schools, or when workers are nested within workplaces. The models provide simultaneous estimates of the differences between individuals, between higher-level units and of the way that those units affect individual differences.
Mixed models can also be used for the analysis of longitudinal data. Applying multilevel models, temporal trajectories, for example a sequence of health measurements over time, are conceptualized as “nested” within individual survey respondents. The shape of the trajectory reveals how an individual's health changes over time, in relation to her or his personal characteristics, such as age, income and family characteristics. Also it is possible to incorporate an additional level of “community” effects. This part of SPIDA will be taught by Professor Georges Monette of York University.
For the lectures and the daily computer lab sessions in SPIDA, we will be using R, an independent open source (i.e., free) statistical software package with wide-ranging pre-programmed statistical procedures and capacity for programming tailored statistical analyses. In addition, R is invaluable for generating informative high-quality graphics. SPIDA begins with a one-day introduction to R by Professor Fox. No previous knowledge of R is expected of participants. A non-profit enterprise based in the research community, R is rapidly becoming an alternative to the major commercial statistical packages for serious data analysis.
The 2012 Program is coordinated by Professor Bryn Greer-Wootten (Professor Emeritus) Department of Geography and Faculty of Environmental Studies, York University, Associate Coordinator, Statistical Consulting Service, and an Associate Director in the Institute for Social Research. Dr. Greer-Wootten was the principal organizer of the SPIDA sessions at York from 2004 to 2011.
The organizing committee members for SPIDA 2012 are drawn from the Institute’s Statistical Consulting Service (SCS), including Professors Robert Cribbie and Michael Friendly [SCS Joint Co-ordinators 2011-2012] and David Flora, all from York’s Department of Psychology, Professor Georges Monette of York’s Department of Mathematics and Statistics, Professor John Fox (Sociology) of McMaster University, and Dr. Hugh McCague, ISR Statistical Consultant. For further information, please see: www.isr.yorku.ca/scs/.