Statistics 701, Fall 1997.

Instructor: Richard Waterman.

Who should take this course?

Anyone whose future job will involve quantitaive analysis or will involve managing or reviewing other people's quantitative work.

Course objectives.

Provide a solid understanding of modern applied statistical modeling techniques.
Enhance skills in communicating quantitative analyses.

Course philosophy.

Most of the interesting and useful ideas in statistics are conceptual rather than mathematical -- therefore this course will focus on understanding, interpreting and communicating modern statistical ideas. The course will be designed to give participants ample opportunity to develop communication and presentation skills (aka written reports and in class presentations).

Topics.

Review of simple and multiple regression.
Exploratory analysis of high dimensional data.
Advanced regression techniques -- focusing on financial and marketing data sets (weighted least squares, robust regression).
Measuring risk with J.P.Morgan's RiskMetrics and CreditMetrics. (Great application of statistics in finance).
Time series.
Classification using logistc regression.
Overview of computer intensive modeling techniques
- Classification and regression trees (CART).
- Neural networks.

Workload.

Five written reports. (Two to three written pages plus graphics).
One in class presentation (which can be group based).

Book.

The Practice of Econometrics. 1991. Ernst R. Berndt. Addison-Wesley. ISBN 0-2-1-17628-9.

Software.

Qu. Why can't we just use JMP?
Ans. It only goes so far -- to implement most modern graphical techniques Splus is a must. If your future job is really going to involve significant quantitative analysis you will very quickly outgrow JMP but it is unlikely you will ever outgrow Splus. Also, JMP has no time series capability whereas Splus does.

The contents of this document are subject to revision. Particularly the choice of additional software to JMP.

Last revised 9/2/97.