Robust estimation using heavy-tailed distributions
The HEAVY package contains routines to perform robust estimation considering heavy-tailed distributions. Currently, the package includes linear regression, linear mixed-effect models, Grubbs’ model, multivariate location and scatter estimation, multivariate regression, penalized splines, random variate generation and some support functions.
- Provide basic functionality for modeling using scale mixtures of normal distributions in R, via a package.
- Calculations associated with parameter estimation are performed by calling routines in C and Fortran.
- Estimation in linear regression, linear mixed effects models, Grubbs’ model, multivariate regression and penalized splines using the EM algorithm.
- Estimation of location and Scatter using multivariate heavy-tailed distributions.
- Implemented families: normal, Cauchy, Student-t, slash and contaminated normal.
- Estimation of the shape parameters for Student-t and slash distributions.
- Multivariate random number generation for the implemented families and the uniform distribution on the p-dimensional sphere.
- Print and summary methods and some sample databases.
Please report any bugs/suggestions/improvements to Felipe Osorio, Universidad Tecnica Federico Santa Maria. If you find these routines useful or not then please let me know. Also, acknowledgement of the use of the routines is appreciated.
Latest binaries and sources for HEAVY are availables from CRAN package repository
To install this package, start R and enter:
Alternatively, you can download the source as a tarball or as a zip file. Unpack this file (thereby creating a directory named, heavy) and install the package source by executing (at the console prompt)
Next, you can load the package by using the command:
The package is provided under the GPL. HEAVY is under active development: new features are being added and old features are being improved. Although the developer will make efforts to preserve backward compatibility, we cannot absolutely guarantee backward compatibility.
To cite the heavy package in publications use:
Osorio, F. (2019). heavy: Robust estimation using heavy-tailed distributions. R package version 0.38.196.\
Some papers using heavy:
- Davie, S., Minto, C., Officer, R., Lordan, C. (2015). Defining value per unit effort in mixed métier fisheries. Fisheries Research 165, 1-10.
- Osorio, F. (2016). Influence diagnostics for robust P-splines using scale mixture of normal distributions. Annals of the Institute of Statistical Mathematics 68, 589-619.
- Singer, J.M., Rocha, F.M.M., Nobre, J.S. (2016). Graphical tools for detecting departures from linear mixed model assumptions and some remedial measures. International Statistical Review 85, 290-324.
About the Author
I’m Assistant Professor at Department of Mathematics of the Universidad Tecnica Federico Santa Maria, Chile.
I contribute as developer/maintainer of the following R packages:
- L1pack - Routines for L1 estimation.
- MVT - Estimation and testing for the multivariate t-distribution.
- SpatialPack - Package for analysis of spatial data.