Sensitivity analysis and assessment of prior model probabilities in MLBMA with application to unsaturated fractured tuff

Ming Ye, Shlomo P Neuman, Philip D. Meyer, Karl Pohlmann

Research output: Contribution to journalArticle

52 Citations (Scopus)

Abstract

[1] Previous application of maximum likelihood Bayesian model averaging (MLBMA, Neuman (2002, 2003)) to alternative variogram models of log air permeability data in fractured tuff has demonstrated its effectiveness in quantifying conceptual model uncertainty and enhancing predictive capability (Ye et al., 2004). A question remained how best to ascribe prior probabilities to competing models. In this paper we examine the extent to which lead statistics of posterior log permeability predictions are sensitive to prior probabilities of seven corresponding variogram models. We then explore the feasibility of quantifying prior model probabilities by (1) maximizing Shannon's entropy H (Shannon, 1948) subject to constraints reflecting a single analyst's (or a group of analysts') prior perception about how plausible each alternative model (or a group of models) is relative to others, and (2) selecting a posteriori the most likely among such maxima corresponding to alternative prior perceptions of various analysts or groups of analysts. Another way to select among alternative prior model probability sets, which, however, is not guaranteed to yield optimum predictive performance (though it did so in our example) and would therefore not be our preferred option, is a minimum-maximum approach according to which one selects a priori the set corresponding to the smallest value of maximum entropy. Whereas maximizing H subject to the prior perception of a single analyst (or group) maximizes the potential for further information gain through conditioning, selecting the smallest among such maxima gives preference to the most informed prior perception among those of several analysts (or groups). We use the same variogram models and log permeability data as Ye et al. (2004) to demonstrate that our proposed approach yields the least amount of posterior entropy (residual uncertainty after conditioning) and enhances predictive model performance as compared to (1) the noninformative neutral case in which all prior model probabilities are set equal to each other and (2) an informed case that nevertheless violates the principle of parsimony.

Original languageEnglish (US)
Article numberW12429
Pages (from-to)1-14
Number of pages14
JournalWater Resources Research
Volume41
Issue number12
DOIs
StatePublished - Dec 2005

Fingerprint

tuff
Sensitivity analysis
sensitivity analysis
entropy
variogram
permeability
Entropy
conditioning
model uncertainty
Air permeability
air permeability
Maximum likelihood
uncertainty
statistics

ASJC Scopus subject areas

  • Environmental Science(all)
  • Environmental Chemistry
  • Aquatic Science
  • Water Science and Technology

Cite this

Sensitivity analysis and assessment of prior model probabilities in MLBMA with application to unsaturated fractured tuff. / Ye, Ming; Neuman, Shlomo P; Meyer, Philip D.; Pohlmann, Karl.

In: Water Resources Research, Vol. 41, No. 12, W12429, 12.2005, p. 1-14.

Research output: Contribution to journalArticle

@article{6e29294b48974fe2b5a31de94cbd115f,
title = "Sensitivity analysis and assessment of prior model probabilities in MLBMA with application to unsaturated fractured tuff",
abstract = "[1] Previous application of maximum likelihood Bayesian model averaging (MLBMA, Neuman (2002, 2003)) to alternative variogram models of log air permeability data in fractured tuff has demonstrated its effectiveness in quantifying conceptual model uncertainty and enhancing predictive capability (Ye et al., 2004). A question remained how best to ascribe prior probabilities to competing models. In this paper we examine the extent to which lead statistics of posterior log permeability predictions are sensitive to prior probabilities of seven corresponding variogram models. We then explore the feasibility of quantifying prior model probabilities by (1) maximizing Shannon's entropy H (Shannon, 1948) subject to constraints reflecting a single analyst's (or a group of analysts') prior perception about how plausible each alternative model (or a group of models) is relative to others, and (2) selecting a posteriori the most likely among such maxima corresponding to alternative prior perceptions of various analysts or groups of analysts. Another way to select among alternative prior model probability sets, which, however, is not guaranteed to yield optimum predictive performance (though it did so in our example) and would therefore not be our preferred option, is a minimum-maximum approach according to which one selects a priori the set corresponding to the smallest value of maximum entropy. Whereas maximizing H subject to the prior perception of a single analyst (or group) maximizes the potential for further information gain through conditioning, selecting the smallest among such maxima gives preference to the most informed prior perception among those of several analysts (or groups). We use the same variogram models and log permeability data as Ye et al. (2004) to demonstrate that our proposed approach yields the least amount of posterior entropy (residual uncertainty after conditioning) and enhances predictive model performance as compared to (1) the noninformative neutral case in which all prior model probabilities are set equal to each other and (2) an informed case that nevertheless violates the principle of parsimony.",
author = "Ming Ye and Neuman, {Shlomo P} and Meyer, {Philip D.} and Karl Pohlmann",
year = "2005",
month = "12",
doi = "10.1029/2005WR004260",
language = "English (US)",
volume = "41",
pages = "1--14",
journal = "Water Resources Research",
issn = "0043-1397",
publisher = "American Geophysical Union",
number = "12",

}

TY - JOUR

T1 - Sensitivity analysis and assessment of prior model probabilities in MLBMA with application to unsaturated fractured tuff

AU - Ye, Ming

AU - Neuman, Shlomo P

AU - Meyer, Philip D.

AU - Pohlmann, Karl

PY - 2005/12

Y1 - 2005/12

N2 - [1] Previous application of maximum likelihood Bayesian model averaging (MLBMA, Neuman (2002, 2003)) to alternative variogram models of log air permeability data in fractured tuff has demonstrated its effectiveness in quantifying conceptual model uncertainty and enhancing predictive capability (Ye et al., 2004). A question remained how best to ascribe prior probabilities to competing models. In this paper we examine the extent to which lead statistics of posterior log permeability predictions are sensitive to prior probabilities of seven corresponding variogram models. We then explore the feasibility of quantifying prior model probabilities by (1) maximizing Shannon's entropy H (Shannon, 1948) subject to constraints reflecting a single analyst's (or a group of analysts') prior perception about how plausible each alternative model (or a group of models) is relative to others, and (2) selecting a posteriori the most likely among such maxima corresponding to alternative prior perceptions of various analysts or groups of analysts. Another way to select among alternative prior model probability sets, which, however, is not guaranteed to yield optimum predictive performance (though it did so in our example) and would therefore not be our preferred option, is a minimum-maximum approach according to which one selects a priori the set corresponding to the smallest value of maximum entropy. Whereas maximizing H subject to the prior perception of a single analyst (or group) maximizes the potential for further information gain through conditioning, selecting the smallest among such maxima gives preference to the most informed prior perception among those of several analysts (or groups). We use the same variogram models and log permeability data as Ye et al. (2004) to demonstrate that our proposed approach yields the least amount of posterior entropy (residual uncertainty after conditioning) and enhances predictive model performance as compared to (1) the noninformative neutral case in which all prior model probabilities are set equal to each other and (2) an informed case that nevertheless violates the principle of parsimony.

AB - [1] Previous application of maximum likelihood Bayesian model averaging (MLBMA, Neuman (2002, 2003)) to alternative variogram models of log air permeability data in fractured tuff has demonstrated its effectiveness in quantifying conceptual model uncertainty and enhancing predictive capability (Ye et al., 2004). A question remained how best to ascribe prior probabilities to competing models. In this paper we examine the extent to which lead statistics of posterior log permeability predictions are sensitive to prior probabilities of seven corresponding variogram models. We then explore the feasibility of quantifying prior model probabilities by (1) maximizing Shannon's entropy H (Shannon, 1948) subject to constraints reflecting a single analyst's (or a group of analysts') prior perception about how plausible each alternative model (or a group of models) is relative to others, and (2) selecting a posteriori the most likely among such maxima corresponding to alternative prior perceptions of various analysts or groups of analysts. Another way to select among alternative prior model probability sets, which, however, is not guaranteed to yield optimum predictive performance (though it did so in our example) and would therefore not be our preferred option, is a minimum-maximum approach according to which one selects a priori the set corresponding to the smallest value of maximum entropy. Whereas maximizing H subject to the prior perception of a single analyst (or group) maximizes the potential for further information gain through conditioning, selecting the smallest among such maxima gives preference to the most informed prior perception among those of several analysts (or groups). We use the same variogram models and log permeability data as Ye et al. (2004) to demonstrate that our proposed approach yields the least amount of posterior entropy (residual uncertainty after conditioning) and enhances predictive model performance as compared to (1) the noninformative neutral case in which all prior model probabilities are set equal to each other and (2) an informed case that nevertheless violates the principle of parsimony.

UR - http://www.scopus.com/inward/record.url?scp=31444456874&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=31444456874&partnerID=8YFLogxK

U2 - 10.1029/2005WR004260

DO - 10.1029/2005WR004260

M3 - Article

AN - SCOPUS:31444456874

VL - 41

SP - 1

EP - 14

JO - Water Resources Research

JF - Water Resources Research

SN - 0043-1397

IS - 12

M1 - W12429

ER -