1999
DOI: 10.1214/aos/1018031101
|View full text |Cite
|
Sign up to set email alerts
|

Coherent dispersion criteria for optimal experimental design

Abstract: We characterize those coherent design criteria which depend only on the dispersion matrix (assumed proper and nonsingular) of the "state of nature," which may be a parameter-vector or a set of future observables, and describe the associated decision problems. Connections are established with the classical approach to optimal design theory for the normal linear model, based on concave functions of the information matrix. Implications of the theory for more general models are also considered.

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
1
1
1

Citation Types

0
89
0
1

Year Published

2012
2012
2019
2019

Publication Types

Select...
8
1

Relationship

0
9

Authors

Journals

citations
Cited by 128 publications
(90 citation statements)
references
References 28 publications
0
89
0
1
Order By: Relevance
“…is proper relative to P (Dawid and Sebastiani, 1999;. For our additive observation model (14,15), the scoring rule…”
Section: Numerical Predictandsmentioning
confidence: 99%
“…is proper relative to P (Dawid and Sebastiani, 1999;. For our additive observation model (14,15), the scoring rule…”
Section: Numerical Predictandsmentioning
confidence: 99%
“…On the other hand, there are scoring rules [e.g., the logarithmic score by Roulston and Smith (2002), applied to a multivariate probability density function] that are more sensitive to misspecified correlations, but require that the forecast is given in terms of a predictive density, and are thus not applicable in the important case of ensemble forecasts. Dawid and Sebastiani (1999) proposed some multivariate scoring rules that depend only on the mean vector m F and the covariance matrix S F of the predictive distribution F. A particularly appealing example is the scoring rule [hereafter referred to as the Dawid-Sebastiani score (DSS)]:…”
Section: Introductionmentioning
confidence: 99%
“…The scoring rules are preferred by [28] to check the predictions from mixed models which include random effects for multivariate modelling. There are three commonly used proper scoring rules: the logarithmic score (LS) [29], the Dawid-Sebastiani score (DSS) [30] and the ranked probability score (RPS) [31, 32]. LS is the negative log-likelihood evaluated at the actual observation; DSS is the standardised difference between the actual observation and predictive mean value plus a penalty of the predictive variance; RPS is the sum of the Brier scores [33] for binary predictions at all possible thresholds.…”
Section: Methodsmentioning
confidence: 99%