STAT 474/574 Biostatical Methods: Survival Analysis and Logistic Regression
Department of Mathematics and Statistics, UNM
Fall Semester 2013
http://www.stat.unm.edu/~ghuerta/stat574/stat574.htm
INSTRUCTOR
Gabriel Huerta
Office: 314 SMLC
email: ghuerta AT stat DOT unm DOT edu
Class Time: Tue-Thurs 12:30-13:45
Classroom: Centenial Engineering Center 1026
Office Hours: T R 14:00-15:15 at SMLC 314 (or by appointment).

DESCRIPTION

An overview of methods commonly used to analyze medical and epidemiological data. Topics include Kaplan-Meier estimate of the survivor function, models for censored survival data, the Cox proportional hazards model, methods for categorical response data including logistic regression and probit analysis, generalized linear models. Methods for correlated data.

TOPICS

  • Linear Models, Generalized Linear Models (GLMs)
  • Estimation of GLMs. Poisson Regression
  • Goodness of fit. Model adequacy.
  • Binary variables. Logistic Regression.
  • Survival data and Survival Functions.
  • Kaplan-Meier and Hazard Estimation.
  • Parametric Survival Models. Weibull model
  • Cox model. Estimation, diagnostics.
  • Correlated data models
  • Bayesian Methods and some MCMC

    COMPUTING ENVIRONMENTS
  • R , and Openbugs/Winbugs . I will provide some basic material and examples in R and Openbugs. I will probably show some examples in SAS too but mostly as a reference. I don't intend to use or teach much SAS this time around. This is not a class on SAS.

    PREREQUISITE

    Stat 428/528 Advanced Data Analysis II or Stat 440/540 Regression Analysis

    TEXTBOOK

  • Dobson, A. and Barnett, A.G. (2008) An Introduction to Generalized Linear Models, Third Editon , Chapman and Hall/CRC.
    • OTHER BOOKS (not required but useful material)

  • The BUGS book . To install Openbugs on a MAC visit this page and follow the instructions there. However the link to wineskin appears broken. Try this link to download Wineskin.
  • Collett, D. (2003) Modelling Survival Data in Medical Research, Second Edition, Chapman and Hall/CRC

    • GRADING

    The grading will be based on homework assignments, a midterm (in-class) and a take home final. The homeworks are worth 30% of the course grade (only selected HW exercises will be graded). The in-class and final take home are each 35%.

    Midterm will be on *Tuesday, October 29th* (in-class) will cover material up to the class of October 24 with some review material on that day. Questions in the test could relate to HW assignments, material in class and relevant textbook sections.

    The take home final will be due on final exams week (date to be determined). Differences between 470/570 will be made mostly for tests.

    Sorry, but I do not accept late homework neither can scheduled tests be at a different time. Exercises from the textbook (Dobson) and other sources as mentioned above will be assigned as HW. I expect HWs to be turned-in as as neatly as you can with clear and concise interpretation of data, analysis, graphs and/or computer output. The final take home will require an analysis of a particular data set to show your proficiency in applying the methods presented in class. The final take home will involve a written report discussing the main results of your data analysis. I prefer you use R and Openbugs for HWs, but SAS will also be allowed. Sorry but I cannot accept HWs or the take home test via e-mail. I prefer you turn in a paper version of your work during class or in office hours at the due date.

    NOTES/HANDOUTS

  • Ch1notes
  • Ch2notes
  • Ch3notes
  • Ch3-Ch4 notes (math details).
  • Notes on Deviance for Binary/Binomial data.
  • Notes on normal linear model.
  • Notes on Ch 8. multinomial logistic regression.
  • Notes Ch. 9 Poisson regression.
  • Notes Survival analysis.
  • Review notes
  • Bayes intro (Binomial distribution). More intro notes on Bayes.
  • Intro to MCMC notes . Gibbs and Metropolis.
    SOME EXAMPLES

  • R code Dobson page 14.
  • Openbugs code Poisson examples.
  • R code for data Table 2.7.
  • Openbugs code for data Table 2.7.
  • Openbugs code simple linear regression example.
  • R code of Exercise 4.1 of Dobson's book
  • Openbugs code related to Ex 4.1
  • R code Newton-Raphson for Poisson distribution (not GLM).
  • R code logistic regression example
  • Table 6.3 data . R code
  • R code Beetle example from Chapter 7.
  • Table 8.1 data R code
  • R code Dobson chapter 9 example. Data file
  • R code , Data and SAS code Collett data page 7.
  • Comparison of 2 groups of survival data R code
  • Fitting parametric survival models Rcode
  • Cox proportional hazards model. Recidivism example data ; R code Description .
  • Bayes calculations for Binomial distribution R code
  • Binomial probability and comparision of proportions Openbugs code
  • Poisson change point example Openbugs
  • More Openbugs examples

    HOMEWORK

    *Will be updated as we move along the semester*

  • HW "zero" (no credit). Try to download R and Openbugs (or Winbugs). Check that you download the book data files.
  • HW No. 1 from Dobson's book. From Chapter 1, exercises 1.1 and 1.6, From Chapter 2, exercise, 2.2 and 2.4. (Due Sept. 5)
  • HW No. 2 Chapter 3 exercises 3.1, 3.3, 3.6, 3.7 and 3.10. Chapter 4, exercise 4.2 (computer exercise). Due date September 17. ( Sept. 19 out of town)
  • HW No. 3 from Dobson's book Chapter 5, exercise 5.2. Chapter 6, Exercises 6.2 and 6.4 Due date September 26.
  • HW No. 4 from Dobson's book Chapter 7, exercise 7.1. Due October 8.
  • HW No. 5 from Dobson's book Chapter 8: 8.2 and Chapter 9: 9.2. Due October 17 (originally). Its fine if you turn it in for October 22nd!
  • *Midterm* October 29th (consider exercises 7.2, 8.1 and 9.1, among others I will suggest for the midterm)
  • HW No. 6 from Dobson's book. Chapter 10: 10.1,10.2 and 10.4. Due November 21.
  • Here is the TAKE HOME PDF version and Word version

    DATA SETS

  • For Dobson's book. Now available as a zip file files
    UNM POLICY ON DISABILITIES

    Qualified students with disabilities needing appropriate academic adjustments should contact me to enures these are met in a timely manner.