[back to main]

Programs

Linked to this page are programs for the following papers. See the publication page for citation details and related papers. The last three programs are specialized software for examples from various papers.

Random partitions with regression on covariates: R package "PPMx"

Applicability: general . The program is suitable for a wide range of applications. it implements the method proposed in Mueller, Quintana and Rosner (2008) (pdf)
The program is implemented as an R library. The algorithm proposed in the paper can be carried out by calling the appropriate functions in the library.
Meaningful defaults are supplied for all tuning parameters. The functions should be straightforward to use for any user who is familiar with R.

Mueller, P., Quintana, F. and Rosner, G. (2008)
"Bayesian Clustering with Regression" (pdf)

Screening Designs for Drug Development: R package 'seqdesphII'

Applicability: general . The program is suitable for a wide range of applications. it implements the method proposed in the Rossell, Mueller and Rosner (2005) (pdf)
The program is implemented as an R library. The algorithm proposed in the paper can be carried out by calling the appropriate functions in the library.
Meaningful defaults are supplied for all tuning parameters. The functions should be straightforward to use for any user who is familiar with R.


A Bayesian Mixture Model for Differential Gene Expression (R and C)

Applicability: general . The program is suitable for a wide range of applications. It implements the method proposed in the paper by Do, Mueller and Tang (2002).

DP Mixture (MDP) of normals: R package 'mdp'

Applicability: general . The program is suitable for a wide range of applications. It implements a popular model for random probability measures (RPM). The RPM would typically be part of a larger probability model, using the RPM, for example, as random effects distribution in a mixed effects model, as exposure model in a retrospective sampling model, as residual distribution in a regression problem, as event time distribution in a survival model, etc.

Description: Implements MCMC for a multivariate DP mixture of normal model for a random probabilty measure.
For a discussion of the model see, for example:

MacEachern, S.N. and Mueller, P. (1998).
``Estimating Mixture of Dirichlet Process Models,''
Journal of Computational and Graphical Statistics , 7, 223--239. (pdf)

Hierarchical DP Mixture: R package 'hdpmn'

Applicability: general . The program is suitable for a wide range of applications. It implements a relatively simple model for a set of dependent random probability measures (RPMs). The RPMs would typically be part of a larger probability model, using the RPMs, for example, as random effects distributions in a mixed effects model across different subpopulations, as sampling models for different subpopulations in a hierarchical model, etc.

Description: Implements MCMC for a set of dependent random probability measures H[j] with a hierarchical DP prior. The hierarchy is constructed by assuming that each H[j] arises as a mixture of a common measure F[0] and a idiosyncratic measure F[j] specific to the submodel.
For details see

Mueller, P., Quintana, F. and Rosner, G. (2004).
``Hierarchical Meta-Analysis over Related Non-parametric Bayesian Models.''
Journal of the Royal Statistical Society, Series B , 66, 735--749. (pdf)

ANOVA DDP (univariate): R package 'ddpanova'

Applicability: general . The program implements the ANOVA DDP model for univariate outcomes. The program is suitable for a wide range of applications. It implements a model for dependent random probability measures (RPM). The nature of the dependence is an ANOVA structure. The RPMs might be sampling models for event times in a survival analysis across related (but not exchangeable) subpopulations. Or the RPMs might be used as random effects distributions in a mixed effects model, etc.

Description: The R package implements the ANOVA DDP model as described in Mueller et al (2005), but without the top level longitudinal profile model.
Now with new function for (censored) event time data.

De Iorio, M., Mueller, P., Rosner, G., and Maceachern, S. (2004). ``An ANOVA Model for Dependent Random Measures,'' JASA, 99(465), 205--215. (pdf)

Hierarchical DP mixture for mixed effects models: R package 'hdpm'

Applicability: specialized . The program is specific to the example in the paper Mueller et al. (2004) for semiparametric inference for repeated measurements data across related subpopulations. The specific model for the longitudinal profile is hard-coded. A user who wishes to adapt the program to another application, with a different parametrization of the longitudinal would need to change some of the underlying C functions. The program includes documentation on how to do this. But it requires some expertise.

Description: This package implements a semi-parametric mixed efffects model for related studies. The random effects distribution is a hierarchical DP mixture as in hdpmn . An additional nonlinear regression defines a sampling model for the repeated measurements over time, conditional on subject-specific random effects. The program implements the specific nonlinear regression model used in the paper. For alternative models, either change the function definitions in the source code (difficult), or use hdpmn with ad-hoc (e.g., m.l.e.) point estimates for the random effects vectors (easy).

Mueller, P., Quintana, F. and Rosner, G. (2004).
``Hierarchical Meta-Analysis over Related Non-parametric Bayesian Models.''
Journal of the Royal Statistical Society, Series B , 66, 735--749. (pdf)

ANOVA DDP (multivariate): R package 'anovaddp'

Applicability: specialized . The program is specific to the example in the paper Mueller et al. (2005) for dependent random probability measures, where dependence is defined in an ANOVA fashion (ANOVA DDP). The program is specialized because it implements the ANOVA DDP in the context of a repeated measurements model with related subpopulations. The ANOVA DDP defines the random effects distributions across subpopulations. The program is specialized because it includes a specific sampling model for the longitudinal measurements (conditional on random effects). It is difficult to adapt the program to general use.
We plan to develop a simplified version of the program to only implement the ANOVA DDP, without the repeated measurements model, that would be useful for other users.

Description: The R package implements the ANOVA DDP model for

Mueller, P., Rosner, G., De Iorio, M., and MacEachern, S. (2005).
``A Nonparametric Bayesian Model for Inference in Related Studies.''
Applied Statistics , 54 (3), 611-626.

SSM: R macros "SSM"

Applicability: illustration example only . These R macros show inference for a proposed new class of models SSM(G0,p) . The implementation serves as illustration example only. It is not suitable for actual data analysis.

Lee, Quintana, Mueller and Trippa (2008)
"Defining Predictive Probability Functions for Species Sampling Models" (pdf)