gamboostLSS: An R Package for Model Building and Variable Selection in the GAMLSS Framework

Abstract

Generalized additive models for location, scale and shape are a flexible class of regression models that allow to model multiple parameters of a distribution function, such as the mean and the standard deviation, simultaneously. With the R package gamboostLSS, we provide a boosting method to fit these models. Variable selection and model choice are naturally available within this regularized regression framework. To introduce and illustrate the R package gamboostLSS and its infrastructure, we use a data set on stunted growth in India. In addition to the specification and application of the model itself, we present a variety of convenience functions, including methods for tuning parameter selection, prediction and visualization of results. The package gamboostLSS is available from the Comprehensive R Archive Network (CRAN) at https://CRAN.R-project.org/package=gamboostLSS.

Modeling spatiotemporal abundance of mobile wildlife in highly variable environments using boosted GAMLSS hurdle models

“…We used the selected sample to estimate the model and the balance of the data in each sample to determine the out‐of‐bag prediction accuracy (empirical risk) measured by the negative log‐likelihood of each model; the optimal stopping iteration () is the iteration with the lowest average empirical risk. In boosted GAMLSS models we used multi‐dimensional subsampling to determine the stopping iteration for each of the GAMLSS parameters while allowing for potentially different model complexities in the parameters; a detailed explanation of this cross‐validation (subsampling) scheme is given in Hofner, Mayr, et al ().…”

“…In boosted GAMLSS models we used multi-dimensional subsampling to determine the stopping iteration for each of the GAMLSS parameters while allowing for potentially different model complexities in the parameters; a detailed explanation of this cross-validation (subsampling) scheme is given in Hofner, Mayr, et al (2016).…”

“…We propose a powerful, iterative modeling approach that combines GAM and GAMLSS in a gradient descent boosting framework (Hofner, Mayr, & Schmid, ; Hothorn, Bühlmann, Kneib, Schmid, & Hofner, ; Mayr, Fenske, Hofner, Kneib, & Schmid, ) to address the challenges of predicting occupancy and abundance from survey data in highly variable environments. Our approach independently evaluates environmental covariates for both occupancy and abundance, while allowing response functions to vary.…”

Boosting joint models for longitudinal and time‐to‐event data

“…The range of the grid (minimum and maximum m stop ) has to be specified ad-hoc, however it might be necessary to adapt it based on the results (adaptive grid search). For a more detailed discussion on how to select this grid in practice, we refer to Hofner et al (2016). Finally, the combination of stopping iterations from the grid is selected, which yields the smallest empirical risk on test-data (e.g., via cross-validation or resampling procedures).…”

“…For a more detailed discussion on how to select this grid in practice, we refer to Hofner et al (2016). Finally, the combination of stopping iterations from the grid is selected, which yields the smallest empirical risk on test-data (e.g.…”