Stat 541 Fall 1999

Instructor

Richard Waterman     waterman@wharton.upenn.edu
http://www-stat.wharton.upenn.edu/~waterman/homepage.html
Tel. 215.573.3637  Fax 215.898.1280
Department of Statistics, 3025 SH-DH, 3620 Locust Walk,
Philadelphia PA 19104.6302.

Objectives

The course will provide students with an exposure to statistical methods intended for exploratory and creative data analysis. Students should become familiar with a range of modern tools and techniques suited to this task. Methods will be implemented in SPlus, the data analysis tools most suited to the approach taken in the course. Statistical theory will be addressed in so far as it guides practice, but not for its own sake.

Resources

Venables & Ripley. (1994) Modern and Applied Statistics with S-Plus, Springer-Verlag.

Myers. (1990) Classical and Modern Regression with Applications, Pws Pub Co.

Frederick Mosteller and John W. Tukey. (1977) Data Analysis and Regression. Addison Wesley. ISBN0-201-04854-X

Hoaglin, Mosteller and Tukey (Eds.) (1982) Understanding robust and exploratory data analysis. Wiley

Stanford Weisberg. (1985) Applied linear regression. Wiley

Cook and Weisberg. (1994) An introduction to regression graphics. Wiley

McCullagh and Nelder. (1989) Generalized Linear Models. Chapman and Hall

Topics

Summarizing, massaging and exploring data

Graphical tools and displays

Numerical summaries

Transformations

Standardization

Scatterplot smoothing

Conditioning/Trellis

The origin of the data -- or the i.i.d. mantra

Sampling

Cluster sampling

Modeling

Classical linear models

Regression

Anova and multiple comparisons

Generalized linear models

Logistic regression

Poisson regression

Model assumptions, diagnostics and graphics

Model selection - cross-validation and out of sample prediction

Variance estimation - Bootstrap and jackknife

Assignments and grading

There will be four assignments over the semester. Each assignment will involve analyzing a dataset, with the purpose of applying techniques learnt in class and interpreting results. Each assignment will contribute 25.00% toward the final grade.