STA 440/540 REGRESSION ANALYSIS

Department of Mathematics and Statistics, UNM
Fall Semester 2006

INSTRUCTOR

Dr. Gabriel Huerta
Office: 441 Humanities Building
email: ghuerta at stat.unm.edu
Phone: 277-2564
Class Time: MWF 15:00-15:50.
Classroom: 234 Dane Smith Hall
Course Web-Page: http://www.stat.unm.edu/~ghuerta/sta540/course.html
Office Hours: MW 12:00-13:00; 16:00-17:00 or by appointment
*Please send me a note by e-mail to make an appointment outside office hours.

DESCRIPTION

This course covers the main aspects of simple linear regression and multiple linear regression. Diagnostics in the form of residual analysis and transformationswill also be considered. The course will discuss a matrix approach to general linear models. Additional topics include model selection procedures, nonlinear least squares and logistic regression. Computer applications mainly using the software MINITAB.

PREREQUISITES

STA 345 , 527 and some background with linear algebra, specifically matrix representations and operations. I will assume familiarity with basic statistical concepts such as distribution function, expectation, variance, correlation, confidence intervals and various probability distributions (Normal, Poisson, Binomial, etc.)

TEXTBOOK (required)

  • Christensen, R. (1996) Analysis of Variance, Design, and Regression: Applied Statistical Methods Chapman and Hall.

    • Basically we will cover: Review Chapter 2 and Chapter 3 with examples from Section 4.2. Chapters 7, 13, and 14, Appendix A, Chapter 15, Sections 16.1.2, 16.3, 16.5 (along with analysis of covariance), Section 8.7 and finally Chapter 18.

    R. Christensen's description on this book with data. LINK

    HOMEWORK

    Will be posted regularly here along the semester.

  • Homework # 1 From Christensen's text exercises 2.7.1,2.7.2, 2.7.3, 2.7.6, 3.7.2, 3.7.3, 3.7.4. Due on Sept. 6.
  • Homework # 2 Prob. 1 For each of the data sets in the Anscombe file, make plots of Y vs. X. Find the simple regression least-squares line. Find the coefficient of determination. From Christensen's text Exercise 7.13.1; 7.13.2; 7.13.3. Due on Sept. 25 (no late HW accepted)
  • Homework # 3 From Christensen's text Chapter 7 Exercises 7.13.5, 7.13.6, 7.13.7, 7.13.11, 7.13.15, 13.8.2,13.8.3 Due on Oct. 18.
  • Midterm Exam (in-class), Monday October 30. This exam will cover material on Ch. 7, sections 7.1-7.11, Ch. 13, sections 13.1-13.7. You may use one page of notes (both sides) and a calculator. Review session on Friday Oct. 27.
  • Solutions for midterm exam.
  • Homework # 4 From textbook Chapter 13, Exercises 13.8.3. Chapter 14, Exercises 14.5.2,14.5.3 and 14.5.5. Exercises 13.8.3 and 14.5.5 can be thought as one exercise. This involves the use of multiple regression, variable selection, diagnostics, etc. Try to present these two exercises as a concise data analysis report. Due on Fri. November 10.
  • Homework # 5 From Chapter 15, page 456. Exercises 15.8.1, 15.8.2 and 15.8.6. From Chapter 8, page 255 Exercise 8.8.9. Due date December 1st.

  • FINAL EXAM pdf . Due on Wednesday Dec. 13 at 4pm.
  • Data set for final TEXT file
  • Forum for students interested in R. Created by Christian Gunning.
    DATA SETS

  • Anscombe data
  • Forbes data
  • Softdrink data
  • Rats data from Weisberg's book.
  • Rats data without case 3.
    HANDOUTS

  • R code for simple linear regression.
  • Variable selection handout
  • Principal Components Regression handout
  • Non-linear regression R code
    SUPPLEMENTAL BOOK (recommended)

  • Weisberg, S. (2005) Applied Linear Regression. Third Edition. Wiley.

    • This book will be used for support material.

    GRADING

    The grading will be based on homework assignments, midterm exam and a final exam/project. Homeworks will be assigned regularly (every other week) and will involve some 'theory' and 'computer' exercises. I expect homeworks to be presented in time (no late HW please) and as neatly as you can, including all relevant information: graphs, detailed proofs,discussion on results, etc. Homework must turn in at class. No e-mail HW will be accepted. The midterm will be an in-class written test. The final is data analysis take home type of project. Regular homework is worth 50% The midterm and final exam each is worth 25% of the course grade.

    SOFTWARE

    Mainly Minitab with some small use of R. No previous computing experience is required.

    Minitab
    Inexpensive rentals
    R software