Survival Data
Analysis in Epidemiology
Biostatistics/Epidemiology 537
Margaret
Sullivan Pepe
Department of Biostatistics
Course Outline
Winter 2002
Lectures: T, Th
Discussion: T
Videotape: Lectures
will be videotaped and available at the Reserve Desk in the Health Sciences
library.
Web
site: Data
files and some course announcements will be on our class specific website
http://courses.washington.edu/b537
Handouts: Extra
copies of the handouts are filed in the Biostatistics (F600) reception area.
Instructor: Margaret
Pepe, Professor
of Biostatistics
Office: HSB F-672
Telephone: 543-1044
e-mail: mspepe@u.washington.edu
Office
hours: Wednesday 11am-1pm Biostatistics conference room F-600 or by appointment
Teaching assistant: Holly Janes, Biostatistics (25%
FTE)
e-mail:
hjanes@u.washington.edu
Laura Lee Johnson
(25%FTE)
e-mail:leejohn@u.washington.edu
.
Required text: Hosmer DW and Lemeshow S . Applied Survival Analysis Wiley,1999
Required software: STATA
Grading: 20% Homework
(distributed on Thursdays, due 1 week later, no late assignments )
20% Class
participation and weekly quiz (10mins on Tuesdays at
20% Midterm 1 (two hour in class on February 7)
20% Project (distributed by February 7, due March 7)
20% Final (March 21, Thursday
using discussion time in exam timetable)
It is assumed that when entering BIOST/EPI 537, you have completed courses in linear regression and logistic regression (Biost/Epi 536) . You should understand the basic statistical concepts of sampling variation, parameter estimation, confidence limits, and statistical hypothesis testing. You should already be familiar with Stata - the computer package used in this course. By the end of this course, you should be able to do the following:
1.
Estimate survival
curves using the Kaplan-Meier estimator
2.
Estimate the
(smoothed) instantaneous mortality or hazard rate
3.
Compare two or more
survival curves using a log-rank or similar test
4.
Fit the Cox model to
survival data and assess the significance, variation (estimates and their
confidence intervals), and scientific impact of the included predictors
5.
Use graphical and
other methods for assessing adequacy of the fitted model.
6.
Use time-dependent
covariates in the Cox model and interpret the coefficients
7.
Use nested cohort
methods and risk-set sampling to approximate a full survival model analysis
8.
Use some simple
methods for analyzing correlated survival data
9.
Construct person-years
at risk tables and analyze them with Poisson regression
10.
Describe the methods
and results to a non-statistical reader.
Much of the
material comes from the lecture notes prepared previously for this course by
Dr. Barbara McKnight, Bill Barlow and Norm Breslow.
Questions from registered students are encouraged. The questions often clarify points on which
several students may share the same uncertainty. If you believe your question is not of
general interest, feel free to ask me your question before or after class. We expect you to talk to your classmates
about the materials and homeworks to gain further
insight.
References
Breslow N and
Day N. Statistical Methods in Cancer Research, Volume 2: The Design and
Analysis of Cohort Studies.
Collett D. Modelling Survival Data in Medical
Research.
Kleinbaum DG. Survival
Analysis: A Self-Learning Text.