Abstract
We present the R package bild for the parametric and graphical analysis of binary longitudinal data. The package performs logistic regression for binary longitudinal data, allowing for serial dependence among observations from a given individual and a random intercept term. Estimation is via maximization of the exact likelihood of a suitably defined model. Missing values and unbalanced data are allowed, with some restrictions. The code of bild is written partly in R language, partly in Fortran 77, interfaced through R. The package is built following the S4 formulation of R methods.
Highlights
This paper describes the R (R Development Core Team 2011) package bild (Goncalves, Cabral, and Azzalini 2012) available from the Comprehensive R Archive Network at http: //CRAN.R-project.org/package=bild for the parametric and graphical analysis of binary longitudinal data
To the generalized estimating equations (GEE) approach when this is applied to binary response data, the present methodology works by introducing a parametric model for the marginal distribution of the response variable but it differs from the GEE approach on another front, in that the parametric analysis developed here is associated to a fully specified stochastic model for the individual profiles
An alternative likelihood-based formulation for a logistic regression model which allows for the presence of serial dependence has been presented by Azzalini (1994); its adaptation to the case of longitudinal data is immediate
Summary
This paper describes the R (R Development Core Team 2011) package bild (Goncalves, Cabral, and Azzalini 2012) available from the Comprehensive R Archive Network at http: //CRAN.R-project.org/package=bild for the parametric and graphical analysis of binary longitudinal data. Important work on the likelihood approach for discrete longitudinal data, with emphasis on the important special case of binary response, has been done by Fitzmaurice and Laird (1993), Fitzmaurice, Laird, and Rotnitzky (1993), and Fitzmaurice, Laird, and Lipsitz (1994), via the so-called mixed-parametrization. Their approach is unsuitable to handle series of different length of response across individuals, and the interpretation of the association parameters is somewhat problematic; see the discussions of these papers. The package is built following the S4 formulation of R methods
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
