Classical Regression Analysis

This course is closely linked with MATH3034.03W, Modern Regression Analysis, for which it is a prerequisite. The emphasis, in contrast to MATH3330.03 and MATH3230.03, will be a more mathematical development of linear models including modern regression techniques. To develop a solid knowledge of regression models, it is strongly advised that you take both MATH3033.03 and MATH3034.03.

The first term will cover matrix formulation of multiple regression, properties and geometry of least square estimation, general linear hypothesis, confidence intervals and regions, multicollinearity, relationship between ANOVA and regression as well as residual analysis. The second term (MATH3034.03) will cover modern regression techniques like diagnostics, crossvalidation, transformations, logistic and Poisson regression, generalized linear models and nonlinear regression models.

Students will use the computer for some exercises. No previous courses in computing are required. The statistical software package SPLUS in a UNIX enrivonment will be used and instructions will be given in class.

The text is R. H. Myers ed., Classical and Modern Regression with Applications(Duxbury).

The final grade may be based (in each term) on assignments, quizzes, a project, one mid-term, and a final examination.

It is advisable that students have taken AS/SC/MATH1132.03 or AS/SC/MATH2570.03 before taking this course.