Lecture:
-
Tuesday & Thursday, 8:15-9:30, 1250 Mayo.
- Instructor: John E. Connett, john-c@biostat.umn.edu, Mayo
A460, 626-3699.
Off-campus: CCBR, Room 200, 2221 Univ. Ave, SE, Mpls., 626-9010.
- Office hours, Mayo A460: Tuesday, Thursday 9:30-10:45 and
by appointment.
Teaching Assistants :
-
Jonathan Johnson: Office hours: Monday 11:00AM - 1:00 PM,
Thursday 1:00-2:00PM, Mayo A-452.
e-mail: jonathaj@biostat.umn.edu
Pei Li: Office hours to be determined. e-mail, peili@biostat.umn.edu
Text:
Geoff Der and Brian Everitt (2002): A Handbook of Statistical Analysis Using SAS, Second Edition,
- Chapman & Hall/CRC, Boca Raton, FL
Main Topics of the Course / Course Description:
A 3-credit course on the use of SAS procedures, intended for students who
know some statistics (e.g., a year from the Biostat 6450-6451 sequence or
a year of statistics from the Statistics Department). There will be an
intro to the SAS data step for students who don't have any SAS experience.
The main part of the course, however, will focus on the use of the
following SAS procedures:
- PROC NPAR1WAY
: Classical nonparametric statistics (Wilcoxon, etc)
- PROC REG
: regression, regression sums-of-squares tables, R-square, F-test,
predictions, etc.
- PROC GLM
: regression/linear models, ANOVA, 'class variables', interaction,
Type I and Type III sums of squares, sequences of models and tests,
multiple comparisons procedures.
- PROC ANOVA
: just an intro to this, since most of it can be accomplished in
PROC GLM
- PROC LOGISTIC
: models, odds ratios, likelihood statistics, AIC,
pseudo-R-square, goodness-of-fit tests, ordered outcomes, hierarchical
modelling, other topics
- PROC GENMOD and CATMOD
: Categorical outcomes, Poisson regression,
loglinear models, repeated measures, GEE (generalized estimating
equations), other topics.
- PROC LIFETEST and LIFEREG
: time-to-event analysis, Kaplan-Meier, testing
group differences in survival, hazard ratios, parametric survival models.
- PROC PHREG
: Survival analysis using Cox regression; covariates,
proportional hazards, modelling, testing the proportional hazards
assumption.
- PROC MIXED
: Repeated measures, longitudinal data, mixed models, inference,
other topics [Note - this is the procedure which is most demanding in
terms of knowledge of stat and linear algebra - may not be included if the
students have less previous exposure.]
- Other Procedures (e.g., LOESS, FACTOR, DISCRIM, PRINCOMP)
may be taught if there
is demand. The class may also include usage of SAS macros and macro variables,
information on the 'Output Delivery System' (ODS) and on formatting output tables.
The basic approach to all the procedures will be to show how they are
used on a variety of health-related datasets to answer specific problems
regarding estimation, testing, or prediction. Several large example
datasets will be used repeatedly throughout the course. There will be a
number of homework assignments. Students will have access to SAS through
a UNIX server in Biostatistics which can be accessed essentially anywhere
on campus, or they can use PC SAS if they have access to that (e.g.,
through the SPH computing lab).
This class has minimal overlap with PubH 6420 (Intro to SAS), PubH
6421 (Advanced Statistical Computing), and the shortcourse offered by the
Division of Epidemiology. It is a semester-long 2-credit course
open to students with minimal prior exposure to SAS but some substantial
prior exposure to statistics.
Learning Objectives:
1. To become familiar with the structure and manipulation of SAS datasets
2. To learn how to use selected SAS procedures to solve specific data
analytic probems.
3. To learn the basic of SAS macros and macro variables
4. To develop skill in the use of SAS graphing methods
Methods of Instruction and Work Expectations:
1. Lectures, computer lab sessions, examples
2. Homework: approximately 10-12 assignments to be handed in and graded
3. Two Exams: one midterm and one final.
Numerous handouts and notes are posted on the course web pages.
More will be added.
Class syllabus:
Class notes for the course: See the
Main Course Web Page.
Grading criteria:
Grades based on percent of total points: