AN OVERVIEW OF COMPOSITE LIKELIHOOD METHODS

By Cristiano Varin, Nancy Reid, David Firth. 2011

1,158 citations recorded by the source corpus.

View Paper: https://www.semanticscholar.org/paper/2c9bb30a136e7970ab1505039222b3ff33ddd7ce

How researchers see this paper

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

Citations in the corpus

Gaussian Copula Marginal Regression (2012, methodology): “An example of simplified estimation procedure is the method of maximum composite marginal likelihood [35, 54].”
A latent variable representation of count data models to accommodate spatial and temporal dependence: application to predicting crash frequency at intersections (2011, background): “The approach is easy to implement and is based on evaluating lower-dimensional marginal probability expressions that do not require simulation (see Bhat et al., 2010a, Varin et al., 2011).”
A Case Study Competition Among Methods for Analyzing Large Spatial Data (2017, methodology): “…to this problem included factoring (1) into a series of conditional distributions (Vecchia 1988; Stein et al. 2004), the use of pseudo-likelihoods (Varin et al. 2011; Eidsvik et al. 2014), modeling in the spectral domain (Fuentes 2007) or using tapered covariance functions (Furrer et al. 2006;…”
On composite likelihood estimation of a multivariate INAR(1) model (2013, background): “In particular, weighting improves on the estimation of the parameters in the mean while unweighted pairwise likelihood is preferable for inference about the association parameters (Varin et al., 2011).”
Econometric Analysis of Vast Covariance Matrices Using Composite Realized Kernels and Their Application to Portfolio Choice (2016, background): “See Varin (2008) and Varin, Reid, and Firth (2011) for a modern review.”
The maximum approximate composite marginal likelihood (MACML) estimation of multinomial probit-based unordered response choice models (2011, background): “…more general class of composite likelihood function approaches (see Lindsay, 1988), is based on maximizing a surrogate likelihood function that compounds much easier-to-compute, lower-dimensional, marginal likelihoods (see Varin, 2008 and Varin et al., 2011 for recent reviews of the CML method).”
Geostatistics of extremes (2012, methodology): “As Varin et al. (2011) describe, composite likelihood is increasingly being used in settings where the likelihood itself is computationally intractable.”
Futures pricing in electricity markets based on stable CARMA spot models (2012, methodology): “This procedure is also theoretically justﬁed by composite likelihood theory (e.g. Varin, Reid and Firth [40], based on arguments going back to Wald [38]), which shows that under quite general regularity conditions the dependence in the likelihood function can be reduced or even ignored.”
Statistical Properties of Covariance Tapers (2013, methodology): “(16) In some circumstances, (E(G))−1 gives the approximate covariance matrix of the estimates obtained from these estimating equations (Varin, Reid, and Firth 2011).”
On the likelihood function of Gaussian max-stable processes (2011, background): “For instance, setting dummy weights to exclude distant pairs for spatially/temporally correlated observations may improve the efficiency of the pairwise composite likelihood (Davis & Yau, 2011; Varin et al., 2011).”
Composite likelihood estimation for the Brown–Resnick process (2013, background): “…= n∑ i=1 ∑ j1< j2 log f (zi, j1 , zi, j2; θ), 3(θ) = n∑ i=1 ∑ j1< j2< j3 log f (zi, j1 , zi, j2 , zi, j3; θ), and the corresponding maximum likelihood estimators θ̂2 and θ̂3, which are consistent and asymptotically Gaussian as n increases (Lindsay, 1988; Cox & Reid, 2004; Varin et al., 2011).”
Modeling Spatial Processes with Unknown Extremal Dependence Class (2017, methodology): “Wefit a bivariate version ofmodel (4), as discussed in Example 3, to the Newlyn oceanographic data analyzed inWadsworth et al. (2017) to illustrate an asymmetric construction, and to compare with the symmetric models fitted therein.”
Non‐stationary intensity‐duration‐frequency curves integrating information concerning teleconnections and climate change (2018, background): “However, given that the second Bartlett identity is valid for each individual likelihood term, computation of the Hessians can be avoided and an alternate sample estimate can be obtained for the sensitivity matrix by (Varin et al., 2011):”
Minimum Scoring Rule Inference (2014, methodology): “In the special case that each Sk is the log score, (8) becomes a (negative log) composite likelihood (see, e.g., Varin et al., 2011).”