Variable Selection : Using Restricted-Set, Multi-Model Presentations
|Starts:||10:00 13 Mar 2013|
|Ends:||12:00 13 Mar 2013|
|What is it:||Seminar|
|Organiser:||School of Arts, Languages and Cultures|
|Who is it for:||Current University students, University staff|
The problem of which variables to select for a particular model is a common one in social science research. Datasets which are multicollinear are often particularly problematic as model selection can be sample specific and unstable. This session explores some of the issues around variable selection and suggests a solution using a general method for modelling based on limited-set, multi-model presentations that show the analytical landscape rather than a single model. It is proposed that this methodology can effectively deal with multicollinearity, provide a more realistic way of representing results and enable research students a considered method for presenting their research.
Although this session uses analyses and graphics from the R and Rcmdr packages, the content is designed be of interest to a general audience.
Travel and Contact Information
Coupland Building 3