R package BayesMultiMode: Bayesian Mode Inference

Nalan Bastürk; Jamie Cross; Peter de Knijff; Lennart Hoogerheide; Paul  Labonne; Herman K. van Dijk

R package BayesMultiMode: Bayesian Mode Inference

Nalan Bastürk, Jamie Cross, Peter de Knijff , Lennart Hoogerheide, Paul Labonne^*, Herman K. van Dijk

^*Corresponding author for this work

Research output: Non-textual / digital / web - outputs › Software › Academic

Abstract

A Bayesian approach for mode inference which works in two steps. First, a mixture distribution is fitted on the data using a sparse finite mixture (SFM) Markov chain Monte Carlo (MCMC) algorithm following Malsiner-Walli, Frühwirth-Schnatter and Grün (2016) <doi:10.1007/s11222-014-9500-2>). The number of mixture components does not have to be known; the size of the mixture is estimated endogenously through the SFM approach. Second, the modes of the estimated mixture at each MCMC draw are retrieved using algorithms specifically tailored for mode detection. These estimates are then used to construct posterior probabilities for the number of modes, their locations and uncertainties, providing a powerful tool for mode inference.

Original language	English
Media of output	Online
Publication status	Published - 2023

Access to Document

https://cran.r-project.org/web/packages/BayesMultiMode/index.html

Cite this

@misc{abf507f29c794a8caa69874dcb1edc44,

title = "R package BayesMultiMode: Bayesian Mode Inference",

abstract = "A Bayesian approach for mode inference which works in two steps. First, a mixture distribution is fitted on the data using a sparse finite mixture (SFM) Markov chain Monte Carlo (MCMC) algorithm following Malsiner-Walli, Fr{\"u}hwirth-Schnatter and Gr{\"u}n (2016) ). The number of mixture components does not have to be known; the size of the mixture is estimated endogenously through the SFM approach. Second, the modes of the estimated mixture at each MCMC draw are retrieved using algorithms specifically tailored for mode detection. These estimates are then used to construct posterior probabilities for the number of modes, their locations and uncertainties, providing a powerful tool for mode inference.",

author = "Nalan Bast{\"u}rk and Jamie Cross and {de Knijff}, Peter and Lennart Hoogerheide and Paul Labonne and {van Dijk}, {Herman K.}",

year = "2023",

language = "English",

}

TY - ADVS

T1 - R package BayesMultiMode: Bayesian Mode Inference

AU - Bastürk, Nalan

AU - Cross, Jamie

AU - de Knijff , Peter

AU - Hoogerheide, Lennart

AU - Labonne, Paul

AU - van Dijk, Herman K.

PY - 2023

Y1 - 2023

N2 - A Bayesian approach for mode inference which works in two steps. First, a mixture distribution is fitted on the data using a sparse finite mixture (SFM) Markov chain Monte Carlo (MCMC) algorithm following Malsiner-Walli, Frühwirth-Schnatter and Grün (2016) ). The number of mixture components does not have to be known; the size of the mixture is estimated endogenously through the SFM approach. Second, the modes of the estimated mixture at each MCMC draw are retrieved using algorithms specifically tailored for mode detection. These estimates are then used to construct posterior probabilities for the number of modes, their locations and uncertainties, providing a powerful tool for mode inference.

AB - A Bayesian approach for mode inference which works in two steps. First, a mixture distribution is fitted on the data using a sparse finite mixture (SFM) Markov chain Monte Carlo (MCMC) algorithm following Malsiner-Walli, Frühwirth-Schnatter and Grün (2016) ). The number of mixture components does not have to be known; the size of the mixture is estimated endogenously through the SFM approach. Second, the modes of the estimated mixture at each MCMC draw are retrieved using algorithms specifically tailored for mode detection. These estimates are then used to construct posterior probabilities for the number of modes, their locations and uncertainties, providing a powerful tool for mode inference.

M3 - Software

ER -