The blupsurv Package


The blupsurv package contains tools for fitting proportional hazards models to clustered univariate or bivariate recurrent event data.

The model captures dependence within clusters and between processes by nested pairs of correlated random effects, specified only through their first two moments. Estimation proceeds by means of an iterative Expectation-Maximization–type algorithm in which the random effects are replaced by their orthodox best linear unbiased predictors, formulated and estimated under an auxiliary Poisson model. This approach avoids the need to specify a parametric frailty distribution.


To install the package directly within R type

install.packages("blupsurv", repos="")

or see the R-forge project page for source and binary downloads.


Users should consult the documentation (PDF), which is also provided within the package.

Developers and statisticians should consult the annotated source code.


Emmanuel Sharef
Robert L. Strawderman