Advertisement

Nonnegatively Constrained Confidence Interval Estimation for Ill-Posed Problems

  • Bert W. Rust
Conference paper

Abstract

Let the m x n linear model
$$\hat{y}=K\bar{x}+\hat{\epsilon}$$
be obtained by discretizing a system of first-kind integral equations
$$\hat{y}_i=\int_{a}^{b}K_i(\xi)x(\xi)d\xi+\hat{\epsilon}_i,i+1,...m,$$
with known functions K i(ξ), unknown function x(ξ) and an m-vector ŷ of measurements corrupted by random errors є drawn from a distribution with zero mean vector and known variance matrix. Consider the problem of estimating linear functions of the form
$$\phi=w^T\bar{x}\approx \int_{a}^{b}w(\xi)x(\xi)d\xi,$$
where w(ξ) is an averaging function designed to elicit some desired information about x(ξ). For such problems, the least squares solution is a highly unstable function of the measurements, and the classical confidence intervals are too wide to be useful. The solution can often be stabilized by imposing physically motivated, a priori non-negativity constraints on x. This paper will show how to extend the classical confidence interval estimation technique to accommodate these nonnegativity constraints and how to compute the resulting much-improved confidence intervals.

Keywords

Duality Theorem Ridge Regression Optimal Interval Confidence Interval Estimation Nonnegativity Constraint 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. [1]
    W. R. Burrus, Utilization of A Priori Information by Means of Mathematical Programming in the Statistical Interpretation of Measured Distributions, Ph.D. Thesis, Ohio State University, ORNL-3743, Oak Ridge National Laboratory, June 1965.Google Scholar
  2. [2]
    W. R. Burrus, B. W. Rust, and J. E. Cope, Constrained interval estimation for linear models with ill-conditioned equations, in Information Linkage between Applied Mathematics and Industry II, A. L. Schoenstadt, et al., eds., Academic Press, New York, 1980, pp. 1–38.Google Scholar
  3. [3]
    C. L. Lawson and R. J. Hanson, Solving Least Squares Problems, Prentice-Hall, Englewood Cliffs, 1974.MATHGoogle Scholar
  4. [4]
    D. P. O’Leary and B. W. Rust, Confidence intervals for inequality-constrained least squares problems, with applications to ill-posed problems, SIAM Jour. Sci. Stat. Comput., 7 (1986), pp. 473–489.MathSciNetMATHCrossRefGoogle Scholar
  5. [5]
    D.L. Phillips, A technique for the numerical solution of certain integral equations of the first kind, J. Assoc. Comput. Mach., 9 (1962), pp. 84–97.MathSciNetMATHGoogle Scholar
  6. [6]
    J. E. Pierce and B. W. Rust, Constrained least squares interval estimation, SIAM Jour. Sci. Stat. Comput., 6 (1985), pp. 670–683.MathSciNetMATHCrossRefGoogle Scholar
  7. [7]
    B. W. Rust and W. R. Burrus, Mathematical Programming and the Numerical Solution of Linear Equations, American Elsevier, New York, 1972.MATHGoogle Scholar
  8. [8]
    B. W. Rust and D. P. O’Leary, Confidence intervals for discrete approximations to ill-posed problems, submitted to Jour. Comp. Graph. Statistics, (in review).Google Scholar
  9. [9]
    Phillip Wolfe, A duality theorem for non-linear programming, Quarterly of Appl. Math., 19 (1961), pp. 239–244.MathSciNetMATHGoogle Scholar

Copyright information

© Springer-Verlag New York, Inc. 1992

Authors and Affiliations

  • Bert W. Rust
    • 1
  1. 1.Center for Computing and Applied MathematicsNational Institute of Standards and TechnologyGaithersburgUSA

Personalised recommendations