Regression models for bivariate count outcomes
Xinling Xu
Department of Epidemiology and Biostatistics
University of South Carolina
Columbia, SC
[email protected]
|
James W. Hardin
Department of Epidemiology and Biostatistics
University of South Carolina
Columbia, SC
[email protected]
|
Abstract. We present a new command, bivcnto, for fitting regression models
suitable for analyzing correlated count outcomes. bivcnto allows
specification of two correlated count outcomes with either two outcome-specific
covariate lists or one common covariate list and fits models using a copula
function approach in the general case or using specific parameterizations by
Marshall and Olkin (1985, Journal of the American Statistical
Association 80: 332–338) or Famoye (2010a, Journal of Applied
Statistics 37: 969–981; 2010b, Statistica Neerlandica 64: 112–124).
bivcnto also calculates a likelihood-ratio test comparing the joint
model with estimation of two independent outcome-specific models.
View all articles by these authors:
Xinling Xu, James W. Hardin
View all articles with these keywords:
bivcnto, copula function, correlated count data, Poisson, negative binomial, Famoye bivariate Poisson regression, Marshall–Olkin bivariate negative binomial regression, Famoye bivariate negative binomial regression, Famoye bivariate generalized Poisson regression, general bivariate count regression
Download citation: BibTeX RIS
Download citation and abstract: BibTeX RIS
|