Non-centred parameterisations for hierarchical models and data augmentation.

By Gareth O. Roberts, Omiros Papaspiliopoulos, M. Sköld. 2003

87 citations recorded by the source corpus.

Subfields: Statistics and Probability, Probabilistic & Bayesian Methods

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Statistics and Probability

In Probabilistic & Bayesian Methods

Citations in the corpus

Efficient Gradient-Based Inference through Transformations between Bayes Nets and Neural Nets (2014, methodology): “ations on the efﬁciency of gradientbased inference. However, the topic of centered vs noncentered parameterizations has been investigated for efﬁ- cient (non-gradient based) Gibbs Sampling in work by Papaspiliopoulos et al. (2003; 2007), which also discusses some strategies for constructing paramete”
Bayesian inference for non‐Gaussian Ornstein–Uhlenbeck stochastic volatility processes (2004, methodology): “Therefore, Papaspiliopoulos et al. (2003) proposed a reparameterization of the hierarchical model, which 'orthogonalizes' the prior structure, in the sense that the missing data are transformed so that they are a priori independent of the parameters.”
Black-Box Variational Inference for Stochastic Differential Equations (2018, unknown): “This requires expressing θ ∼ q ( · ; φ ) as a non-centred param-eterisation (Papaspiliopoulos et al., 2003).”
Particle gibbs with ancestor sampling (2014, background): “This can result in a significant improvement in mixing of the Gibbs sampler, in particular when there are strong dependencies between the system state the parameters [41,51].”
Retrospective Markov chain Monte Carlo methods for Dirichlet process hierarchical models (2007, methodology): “ Neal (2000) , and Algorithm 8 of Neal (2000), where we use three auxiliary states.”
Inference for Lévy‐Driven Stochastic Volatility Models via Adaptive Sequential Monte Carlo (2011, background): “Indeed, in the case of the first publication a relatively general reparameterization scheme, non-centred parameterizations (Papaspiliopoulos et al., 2003) are adopted.”
Scaling limits for the transient phase of local Metropolis–Hastings algorithms (2005, result): “For instance, Roberts et al. (1997) and Roberts and Rosenthal (1998) only prove results rigorously when 7r(-) has components which are independent and identically distributed (at least asymptotically), though subsequent theoretical and empirical work shows that these results apply somewhat more gene”
Analysis of the Gibbs Sampler for Hierarchical Inverse Problems (2013, methodology): “se limit, when Nis large. We have considered the interweaving method of [41], which combines in each iteration centered and non-centered draws of , and the partially non-centered parametrizations of [33], in which the prior is reparametrized as u= t2v t where v t ˘N(0;t 1C 0), for some t2[0;1]. Our”