Textbook for Statistical Methods in Epidemiology

[Robert Wolfe]

What books are useful for teaching a third level statistics course for
epidemiology students? These students have had intro biostatistics (up to
simple linear regression) and multiple linear regression. This third course
covers methods more commonly used in epidemiology, such as logistic,
Poisson, and proportional hazards regression, case-control studies,
proportionate mortality analysis, etc. I have always used Breslow and Day,
but the students find it to be too difficult to work with.

[Peter B. Imrey]

Bob Wolfe asks for recommendations for a third-level text on epidemiologic
statistics, indicating that Breslow and Day has proven too difficult for his
students. I have previously tried three texts in this area: Breslow and Day
Vol. 1, with some material from Vol. 2; Kleinbaum, Kupper and Morgenstern;
and Fleiss' Statistical Methods for Rates and Proportions, with the latter
two used in combination. My experience with Breslow and Day was similar to
Bob's, nor was my experience with the other two particularly satisfactory.
All these books are excellent, but both BD and KKM are written with a flavor
that assumes too much practical experience and familiarity with the
conceptual issues involved to be comfortable for students who are just
beginning to get the feel of things. So the development comes across as
less orderly and concise than most students expect, and leaves many feeling
at sea. On the other hand, Fleiss and other statistical texts do not
themselves provide the breadth and depth of epidemiologic focus that is most
desireable, or full coverage of some important statistical issues. There
are many epidemiologic and statistical texts that can be very useful in such
a course, but I've yet to find one which stands alone and fills the bill
without producing more anxiety in the students than seems necessary --
although possibly a highly talented instructor could alleviate this anxiety
with the proper planning. Selvin's book is interesting, but the coverage
doesn't fully match what I like to teach. Some people have used
Schlesselman successfully, but the focus on case-control studies seems too
limiting to me. I'd be very much interested in other responses to Bob's
query.
=====================================================================
Peter B. Imrey, Prof. of Biometrics & Statistics, Univ. of Illinois
Depts. of Medical Information Science, Community Health, & Statistics
506 S. Mathews Ave., Urbana, IL 61801-3618
Ph. (217) 333-2427; FAX (217) 333-8868
p-i...@uiuc.edu, im...@vmd.cso.uiuc.edu, im...@uiucvmd.bitnet