Data
Smoking DATA (ascii data format): Longitudinal binary data on smoking analyzed in Preisser, Galecki, Lohman and Wagenknecht (2000, JASA 95, 1021-1031)
Urinary Incontinence DATA (ascii data format): Correlated binary data relating to urinary incontinence in elderly patients in primary care medical practices analyzed in Preisser and Qaqish (1999, Biometrics 55, 574-579).
Health Maintenance Visit DATA (ascii and SAS (zipped) data formats): Multi-level binary data (patients nested within physicians nested within medical practices) relating to whether a patient made at least one ``health maintenance visit" during the years 1990 and 1991. This data, from the North Carolina Early Cancer Detection Program at the Lineberger Comprehensive Cancer Center, was used to illustrate deletion diagnostics for GEE (Preisser and Qaqish 1996, Biometrika 83:551-562) and for alternating logistic regressions (Preisser, By, Perin and Qaqish 2012, Biometrical Journal 54:701-715).
Clustered Binary Outcome from a Cluster Trial to Combat Underage Drinking (ascii and SAS (zipped) data formats): The Enforcing Underage Drinking Laws (EUDL) program employed a parallel cross-sectional non-randomized cluster trial design with three time points in three rounds. At the time of its implementation, EUDL was the largest federal initiative within the United States to combat underage drinking. States administered funds to selected communities to reduce underage drinking. In the evaluation of the program led by Mark Wolfson (PI: Wake Forest University), propensity score matching was used to select control communities (Preisser et al. 2003) for matching with intervention communities. A variety of outcomes were measured using telephone surveys of youth, ages 16-20. The dataset provided here consists of baseline and year 1 follow-up data from 1346 youths in 38 communities in three states (Michigan, Ohio and Wisconsin) in Round 1. Cluster sizes range from 27 to 41 with mean size 35.4. Three papers illustrate GEE-type analysis (Lu et al., 2007; Preisser and Perin, 2007; Preisser et al., 2008) with pairwise correlations for within-cluster associations whereas Perin and Preisser (2017) use pairwise odds ratios; all assess the effect of the community-level intervention on the binary outcome self-reported last 30-day alcohol use. Extensive other data from the EUDL study are available via data use agreement