John Preisser's Personal Home Page

John Preisser's University of North Carolina Page


This is a brief overview. No warranty or guarantee of any kind is offered with this software. The software can be freely used for non-commercial purposes and can be freely distributed for non-commercial purposes only. For more information or to download the software, click on individual items.

ALR: The SAS PROC GENMOD code used to obtain alternating logistic regression results from the green tobacco sickness data presented in Preisser, Arcury, and Quandt (2003, Am J Epidemiol, 158:495-501) is provided. Program statements are provided for the 3 level model presented in Tables 3 and 4 of the paper. Also, code for a more complicated 4-parameter model (expression 6 in the paper) making use of the ZDATA option on the repeated statement of SAS PROC GENMOD is provided. Alternatively, the sas program provides a simple example showing how SAS PROC SQL may be used to create the Z-matrix for the log pairwise odds ratio model.

GEEORD: A SAS macro GEEORD, as described by Gao X, Schwartz T, Preisser JS, and Perin J, for the analysis of ordinal responses with repeated measures through a regression model that flexibly allows the proportional odds assumption to apply (or not) separately for each explanatory variable. The macro additionally provides relevant tests of the proportional odds assumption.
Version 1.02 by John Preisser and Jamie Perin. Modifications in Version 1.03 by Todd Schwartz (June 10, 2017).

GEEDIAG: A SAS macro GEEDIAG, originally described by Hammill & Preisser (Comput. Statist Data Analysis 2006;51:1197-1212), applies generalized estimating equations (Liang & Zeger, 1986) for estimation of population-averaged models. It includes the following in addition to "standard" options: (1) GEE1 regression diagnostics based on Preisser & Qaqish (1996, Biometrika 83, 551-62) inlcuding DFBETA and Cook's Distance based upon cluster and observation deletion; (2) bias-corrected covariance estimates for marginal mean regression parameters (Mancl & DeRouen, 2001, Bioometrics, 57, 126-134); (3) time-variant, heterogeneous scale parameter specification; and (4) several options to produce output SAS datasets for post-modeling calculations.
Version 1.02 by Bradley Hammill and John Preisser (2005); Version 1.05 by John Preisser (2018).

GEECORR: GEECORR SAS macro for binary data, logit link for marginal mean model, and identity link for linear correlation model based upon a generalized estimating equations procedure of Prentice (1988, Biometrics, 44, 1033-1048). The macro provides estimates of parameters and their standard errors for regression coefficients in the marginal mean model and in the within-cluster pairwise correlation model. It provides bias-corrected covariance estimates that extend those of Mancl & DeRouen (2001, Biometrics, 57, 126-134) to encompass correlation parameters as well as mean parameters. The macro also provides deletion diagnostics for clusters and observations (DBETA, Cook's Distance) proposed by Preisser and Perin (Statistics and Computing, 2007, 17, 381-393) that extend the GEE1 diagnostics of Preisser and Qaqish (1996, Biometrika, 83, 551-562) to include correlation model parameters. Version 1.04 was posted 1-24-07.
Version 1.01 by Richard Zink (2003). Modifications in Version 1.04 by Jamie Perin and John Preisser.

Marginalized Mixture Models for Count Outcomes: A file containing supplementary material, including SAS Proc NLMIXED code for fitting the models as well as information on selection of starting values for parameters, for the article "Marginalized mixture models for count data from multiple source populations" by Habtamu K. Benecha, Brian Neelon, Kimon Divaris and John S. Preisser, published in the Journal of Statistical Distributions and Applications (2017), 4:3.

ORTH/ALR regression diagnostics (link to location on B. Qaqish's software page): The SAS/IML macro applies alternating logistic regressions or (optionally) a generalization of the procedure called orthogonalized residuals (Qaqish, Zink, Preisser, Scandinavian Journal of Statistics, 2012; 39:515-527) and it produces cluster-deletion diagnostics (e.g., Cooks Distance, DFBETA) for both the marginal mean (logistic) model and the within-cluster association (log odds ratio) model (Preisser, By, Perin, Qaqish, Biometrical Journal, 2012; 54:701-715). By Kunthel By, John Preisser, Jamie Perin, Richard Zink and Bahjat Qaqish.

Go To Bahjat F. Qaqish's Software Page

last modified June 13, 2017