eiPack: Ecological Inference and Higher-Dimension Data
Management Olivia Lau, Ryan Moore, and Mike Kellermann Version
0.1-3 (beta)
Ecological inference (EI) takes contingency tables as the unit of
analysis. These tables are described by row and column totals (or
proportions); the goal of inference is to determine the marginal
intra-table relationship between the rows and columns. While some
existing R packages (MCMCpack by Andrew Martin and Kevin
Quinn and eco by Kosuke Imai and Ying Lu, for example) offer
functions that analyze 2 x 2 models, we implement more general methods
that can take more than two rows or columns. Our package
includes:
Extreme case analysis, or the method of
bounds, suggested by Duncan and Davis (1953)
Ecological regression described in Goodman (1953) using both
frequentist point estimates and a Bayesian estimator that
produces correct standard errors.
R x C model described in Rosen et al. (2001) using a Bayesian Markov-chain
Monte Carlo algorithm
Lau, Olivia, Ryan T. Moore, and Michael Kellermann. "eiPack: Ecological
Inference and Higher-Dimension Data Management," R News,
vol.7, no. 2 (October 2007).