[R-pkgs] DPpackage - New version
Alejandro Jara Vallejos
Alejandro.JaraVallejos at med.kuleuven.be
Tue May 29 18:10:27 CEST 2007
I have uploaded version 1.0-4 of DPpackage on CRAN. Since the first
version (1.0-0), I have not communicated the improvements of the
package. I'll use this email to summarize its current status.
The name of the package is motivated by the Dirichlet process.
However, DPpackage tries to be a general package for Bayesian
nonparametric and semi-parametric data analysis. So far, the package
includes models based on Dirichlet processes, Dirichlet process
mixtures of normals, Polya trees, and Random Bernstein polynomials. A
list of current functions is given next:
1) Density estimation: DPdensity (using DPM of normals), PTdensity
(using Mixtures of Polya Trees), and BDPdensity (using
Bernstein-Dirichlet prior). The first two functions allow uni- and
2) Nonparametric random effects distributions in mixed effects models:
2.1) DPlmm and DPMlmm, using a DP/MDP and DPM of normals prior,
respectively, for the linear mixed effects model.
2.2) DPglmm and DPMglmm, using a DP/MDP and DPM of normals prior,
respectively, for generalized linear mixed effects models,
respectively. The sampling(link) considered by these functions are
binomial(logit,probit), poisson(log) and gamma(log).
2.3) DPolmm and DPMolmm, using a DP/MDP and DPM of normals prior,
respectively, for the probit-ordinal mixed effects models.
2.4) DPrasch and FPTrasch, using a DP/MDP and finite PT/MPT
(mixture of Polya Trees) prior for the Rasch model with binary
sampling distribution, respectively.
2.5) DPraschpoisson and FPTraschpoisson. The same as before (2.4)
but with a Poisson sampling.
2.6) DPmeta and DPMmeta for the random (mixed) effects
meta-analysis models, using a DP/MDP and DPM of normals prior,
3) Binary regression with nonparametric link:
3.1) CSDPbinary, using Newton, Czado and Chappell (1996)'s
centrally standardized DP prior.
3.2) DPbinary, using the regular DP prior for the inverse of the
3.3) FPTbinary, using a finite PT prior for the inverse of the
4) AFT model for interval-censored data:
4.1) DPsurvint, using a MDP prior for the baseline distribution.
5) ROC curve estimation:
5.1) DProc, using DPM of normals.
6) Linear model with a nonparametric for the error distribution:
6.1) PTlm, using MPT.
7) DP prior elicitation:
7.1) DPelicit, using the exact and approximated formulas for the
mean and variance of the number of clusters given the total mass
parameter and the number of subjects.
Tim Hanson and Fernando Quintana have made contributions to the
current version. I would also like to thank George Karabatsos for his
input to the current status of the package and Peter Mueller for
actively promoting the package.
Various other improvements have been motivated by questions asked by
many people around the world. I would like to thank all of them too.
I welcome anyone who sends comments, suggestions, remarks, and
particularly those who find bugs or mistakes in any part of the
package or its documentation. DPpackage is an open source program for
Bayesian nonparametric developments. All contributions are welcome.
More information about the R-packages