Witte Lab          

Departments of Epidemiology & Biostatistics, and Urology
University of California, San Francisco

Home
Research
Lab Members
Software
Teaching & Textbook
Biosketch
Links
Personal Info                                             

Last updated: September 11, 2003
©  2002

Software for Hierarchical Modeling of Epidemiologic Data 

The purpose of this site is to distribute SAS IML code for epidemiologic analysis by two-stage semi- or empirical-Bayes hierarchical modeling.  This site complements the manuscript "Software for hierarchical modeling of epidemiolgic data" (Witte et al., Epidemiology 1998;9:563-566).  In particular, we expand on the paper's Appendix, providing SAS IML code corresponding to the diet and breast cancer application first presented in Witte et al. (Epidemiology 1994; 5:612-621).

To use this code, one inputs conventional coefficient and covariance-matrix estimates (e.g., from any logistic software package) and second-stage data (i.e., the second-stage design matrix Z, and, if semi-Bayes, pre-specified second-stage variance values).   As currently written, the code allows one to reproduce our diet and breast cancer hierarchical modeling analysis.  Hence, in addition to the code, we give the conventional logistic regression coefficients and corresponding covariance matrix, as well as the second-stage design matrix and second-stage variance (tau^2) values.

To undertake hierarchical modeling with this SAS code and data, follow these steps:

1. Put a floppy disk into your computer's a: drive.
2. Save the code "hm_sas2.txt" to the floppy disk.
3. Save the first-stage coefficient estimates "bb.txt," the first-stage covariance matrix "bbcov.txt," the second-stage design matrix "zmatrix2.txt," the second-stage variance values "tau2.txt," and a vector of unit increases "s.txt" to the floppy disk.   (Note: to save the SAS and txt files, open them with a click, pull down the file menu, and save.)
4. Open SAS (on your own system), and run the program.