skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: Selecting an Informative/Discriminating Multivariate Response for Inverse Prediction.


Abstract not provided.

; ;  [1];  [1];  [1];  [2]
  1. (LANL)
  2. (SNL)
Publication Date:
Research Org.:
Sandia National Lab. (SNL-NM), Albuquerque, NM (United States)
Sponsoring Org.:
USDOE National Nuclear Security Administration (NNSA)
OSTI Identifier:
Report Number(s):
DOE Contract Number:
Resource Type:
Resource Relation:
Conference: Proposed for presentation at the Joint Statistical Meetings held July 30 - August 4, 2016 in Chicago, IL, United States.
Country of Publication:
United States

Citation Formats

Thomas, Edward V., Lewis, John R, Anderson-Cook, Christine, Burr, Tom, Hamada, Michael S., and Zhang, Adah. Selecting an Informative/Discriminating Multivariate Response for Inverse Prediction.. United States: N. p., 2016. Web.
Thomas, Edward V., Lewis, John R, Anderson-Cook, Christine, Burr, Tom, Hamada, Michael S., & Zhang, Adah. Selecting an Informative/Discriminating Multivariate Response for Inverse Prediction.. United States.
Thomas, Edward V., Lewis, John R, Anderson-Cook, Christine, Burr, Tom, Hamada, Michael S., and Zhang, Adah. 2016. "Selecting an Informative/Discriminating Multivariate Response for Inverse Prediction.". United States. doi:.
title = {Selecting an Informative/Discriminating Multivariate Response for Inverse Prediction.},
author = {Thomas, Edward V. and Lewis, John R and Anderson-Cook, Christine and Burr, Tom and Hamada, Michael S. and Zhang, Adah},
abstractNote = {Abstract not provided.},
doi = {},
journal = {},
number = ,
volume = ,
place = {United States},
year = 2016,
month = 7

Other availability
Please see Document Availability for additional information on obtaining the full-text document. Library patrons may search WorldCat to identify libraries that hold this conference proceeding.

Save / Share:
  • The inverse prediction is important in a variety of scientific and engineering applications, such as to predict properties/characteristics of an object by using multiple measurements obtained from it. Inverse prediction can be accomplished by inverting parameterized forward models that relate the measurements (responses) to the properties/characteristics of interest. Sometimes forward models are computational/science based; but often, forward models are empirically based response surface models, obtained by using the results of controlled experimentation. For empirical models, it is important that the experiments provide a sound basis to develop accurate forward models in terms of the properties/characteristics (factors). And while nature dictatesmore » the causal relationships between factors and responses, experimenters can control the complexity, accuracy, and precision of forward models constructed via selection of factors, factor levels, and the set of trials that are performed. Recognition of the uncertainty in the estimated forward models leads to an errors-in-variables approach for inverse prediction. The forward models (estimated by experiments or science based) can also be used to analyze how well candidate responses complement one another for inverse prediction over the range of the factor space of interest. Furthermore, one may find that some responses are complementary, redundant, or noninformative. Simple analysis and examples illustrate how an informative and discriminating subset of responses could be selected among candidates in cases where the number of responses that can be acquired during inverse prediction is limited by difficulty, expense, and/or availability of material.« less
  • A number of interesting problems in the design of experiments such as sensor allocation, selection of sites for the observing stations, determining sampler positions in traffic monitoring, and which variables to survey/measure in sampling studies may be considered in the following setting: Given a covariance matrix of multi-dimension random vector and given a ratio of the number of possible observations to the observational error select those components which must be observed to guarantee minimization of an objective function describing the quality of prediction of all or prescribed components. The authors show that the problem can be considered in the frameworkmore » of convex design theory and derive some simple but effective algorithm for selection of an optimal subset of components to be observed.« less
  • Abstract not provided.
  • Highly accurate multiple linear regression models have been developed for sandstones of diverse compositions. Porosity reduction or enhancement processes are controlled by the fundamental variables, Pressure (P), Temperature (T), Time (t), and Composition (X), where composition includes mineralogy, size, sorting, fluid composition, etc. The multiple linear regression equation, of which all linear porosity prediction models are subsets, takes the generalized form: Porosity = C{sub 0} + C{sub 1}(P) + C{sub 2}(T) + C{sub 3}(X) + C{sub 4}(t) + C{sub 5}(PT) + C{sub 6}(PX) + C{sub 7}(Pt) + C{sub 8}(TX) + C{sub 9}(Tt) + C{sub 10}(Xt) + C{sub 11}(PTX) + C{submore » 12}(PXt) + C{sub 13}(PTt) + C{sub 14}(TXt) + C{sub 15}(PTXt). The first four primary variables are often interactive, thus requiring terms involving two or more primary variables (the form shown implies interaction and not necessarily multiplication). The final terms used may also involve simple mathematic transforms such as log X, e{sup T}, X{sup 2}, or more complex transformations such as the Time-Temperature Index (TTI). The X term in the equation above represents a suite of compositional variable and, therefore, a fully expanded equation may include a series of terms incorporating these variables. Numerous published bivariate porosity prediction models involving P (or depth) or Tt (TTI) are effective to a degree, largely because of the high degree of colinearity between p and TTI. However, all such bivariate models ignore the unique contributions of P and Tt, as well as various X terms. These simpler models become poor predictors in regions where colinear relations change, were important variables have been ignored, or where the database does not include a sufficient range or weight distribution for the critical variables.« less
  • Remote spectral sensing offers an attractive means of mapping river water quality over wide spatial regions. While previous research has focused on development of spectral indices and models to predict river water quality based on remote images, little attention has been paid to subsequent validation of these predictions. To address this oversight, we describe a retrospective analysis of remote, multispectral Compact Airborne Spectrographic Imager (CASI) images of the Ohio River and its Licking River and Little Miami River tributaries. In conjunction with the CASI acquisitions, ground truth measurements of chlorophyll-a concentration and turbidity were made for a small set ofmore » locations in the Ohio River. Partial least squares regression models relating the remote river images to ground truth measurements of chlorophyll-a concentration and turbidity for the Ohio River were developed. Employing these multivariate models, chlorophyll-a concentrations and turbidity levels were predicted in river pixels lacking ground truth measurements, generating detailed estimated water quality maps. An important but often neglected step in the regression process is to validate prediction results using a spectral residual statistic. For both the chlorophyll-a and turbidity regression models, a spectral residual value was calculated for each river pixel and compared to the associated statistical confidence limit for the model. These spectral residual statistic results revealed that while the chlorophyll-a and turbidity models could validly be applied to a vast majority of Ohio River and Licking River pixels, application of these models to Little Miami River pixels was inappropriate due to an unmodeled source of spectral variation.« less