# Statistics 420

STAT 420: Modern Regression Analysis

Prerequisites: STAT 302

Credit Hours: (3)

Fundamental concepts in modern regression diagnostics. Choice of best subset model using cross validation methods. Study of collinearity and methods for combating collinearity. Use of residuals in regression diagnostics.  Detection of high influence data points.  Use of modern computer software.

Detailed Description of Content of the Course

This course is geared toward building "best" regression models using real world data. The latest statistical techniques in regression analysis are discussed and then applied to various kinds of data situations. The students learn the pitfalls of drawing conclusions about "best model" too quickly by first hand experience.

The following topics are covered:

a) Review of Simple Linear Regression
b) Multiple Linear Regression
c) Criteria for choice of Best Model
d) Analysis of Residuals
e) Influence Diagnostics
f) Detecting and Combatting Mulicollinearity

If time permits the following topics are also covered:

a) Logistic Regression
b) Nonlinear Regression

Detailed Description of Conduct of Course

The course emphasizes solving real world problems using the latest in regression techniques. The lectures are used to present the theory behind the statistical techniques being taught and then illustrations to show the effectiveness.

The students are asked to do a lot of homework. Some of the homework is to help them appreciate the theory behind the techniques they will be using. Most of the homework, however, is problem solving.

These problems are given to the student in the form of data that has been collected. They are asked to analyze the data using the techniques they know at that point and report their conclusions in a form the subject matter scientists (the person responsible for the data) could understand.

Some of the work is done alone and some is done in groups. It is expected of each student that he or she draws conclusions independently.

In class tests are given to encourage the students to become more familiar with the formula they are using and to understand when they are appropriate.

Goals and Objectives of the Course

The goals are:

• to provide the students an opportunity to use the mathematics they have learned in prior courses in practical, real life data analysis.
• to provide students with the tools needed to analyze the large volume of real world problems that fall into the category of "regression analysis."
• to expose students to the usefulness of statistics, particularly in the area of regression analysis.
• to help the students to fully understand and be able to explain the results of their analysis.

Assessment Measures

The means used to assess the students knowledge and ability to apply the course material are in the form of:

• Quizzes in class
• Tests in class
• Homework problems
• Projects to find "best regression model" on a certain set of data
• Final Exam

Other Course Information

This course is computer intensive and as such provides great experience in using statistical software.

Review and Approval
Sept. 2001 Review Stephen Corwin, Chair