M467 | 26213 | Michael Trosset

This course will provide a careful study of linear regression, balancing discussion of the statistical principles (e.g., the method of least squares) that underlie regression methodology with consideration of computational and diagnostic concerns. An important theme will be the use of state-of-the-art graphical methods that facilitate interpretation of regression analyses. Although no previous knowledge of regression is assumed, the treatment of topics such as weighted regression, lack of fit, polynomial regression, data transformation, the analysis of residuals, outliers and influence, and variable selection will be in sufficient depth that many students may find this to be a useful second course on regression. More generally, this would be an excellent second course in statistics for undergraduates who anticipate jobs and/or graduate study in quantitative disciplines. Prerequisites: Students should have already taken at least one course in statistics, have some facility with matrix algebra, and have some experience in writing simple computer programs. (Note: these requirements supersede the generic catalog requirements for M467of M365 and M366; in particular, M366 is not a prerequisite for this course.) If you are uncertain whether or not your background is adequate, then please contact the instructor, Michael Trosset, for more information and obtain his consent. Texts: Applied Linear Regression, by Sanford Weisberg. Publisher: Wiley- Interscience; 3 edition. ISBN: 0471663794 Applied Regression Including Computing and Graphics, by R. Dennis Cook and Sanford Weisberg. Publisher: Wiley-Interscience. ISBN: 047131711X