BayesianRecordLinkage.jl: Julia implementation of unsupervised methods for one-to-one record linkage. Both an EM algorithm based method and a MCMC algorithm employing a standard Fellegi-Sunter based approach are included. The module also includes methods for post-hoc blocking and a penalized likelihood based point estimate. See McVeigh, Spahn, and Murray (2019) for details.
AssignmentSolver.jl: Julia implementation of algorithms for solving linear sum assignment problems. This includes a version of the well known Hungarian algorithm as well as Auction algorithms (Bertsekas 1992).
Software for inferring heterogeneous treatment effects using Bayesian Causal Forests; see Hahn, Murray and Carvalho (2017)
The stable version is available on CRAN. Note that versions prior to 1.2 had a bug in the interface and should be updated. See https://github.com/jaredsmurray/bcf for the latest updates
MixedDataImpute is an R package for imputing missing multivariate continuous and categorical data, using the Bayesian nonparametric hierarchical model developed in Murray and Reiter (2016)
The latest stable version is available from CRAN (“install.packages(MixedDataImpute)”).
bfa is an R package for fitting Bayesian factor models (Gaussian, probit, mixed and semiparametric copula models) under a range of priors for the factor loadings. These include continuous and spike-and-slab shrinkage priors. MCMC is performed in compiled code, so model fitting is typically quite fast.
The latest stable version is available from CRAN (“install.packages(‘bfa’)”).
Here you can find the MPTuner software, an implementation of adaptive median polish algorithms for anomaly detection in sensor networks. The algorithms were developed by Prof. Ernst Linder at the University of New Hampshire’s Department of Mathematics and Statistics as part of the 2007 SAMSI program on Environmental Sensor Networks. Dr. Zoe Cardon at Woods Hole provided us with test data and insight into the ecological processes involved. I wrote the software during my senior year under the mentorship of Prof. Linder. For background on the technique you may refer to this talk by Prof. Linder.
The software is available as the original Python source or as a Windows executable. The scripts should run anywhere that Python and wxPython do. The binary is for Windows platforms only and has been tested in XP, Vista and Windows 7.
The data included here are derived from work supported by the National Science Foundation under Grant No. 0415938. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation.