Clustered data methods
- Current research with You-Gan Wang on efficiency
impacts of working correlation
misspecification (Wang and Carey, Biometrika 2003, JASA 2004)
and on improved methods for
working correlation estimation
- Contributions to computations for flexible working correlation modeling
in the current yags (yet another GEE solver) distribution for R
- Highly general models for clustering in binary data are
supported by alternating logistic regression (Carey, Zeger, Diggle,
Biometrika 1993)
- New methods for calibrated outlier detection for exchangeable multivariate
Gaussian outcomes (Carey and Rosner, submitted)
- A new model for twin-twin correlation that allows correlation
to vary smoothly
as a function of twin-twin discrepancy in exposure and
mono-/di-zygosity of twinship (Tishler, Carey, Reed, Fabsitz; Genetic Epidemiology, 2001)
- A new model for bivariate longitudinal data with
irregular observation times and parsimonious parametrization
of cross-component correlation (Carey and Rosner, Stat in Med 2001)
http://www.biostat.harvard.edu/~carey
