Statistical Engineering Division
Seminar Series
Calculating the Area Under a Curve Given by a General Set of Points
Maurice Cox
National Physical Laboratory, UK
Statistical Engineering Division, NIST
Lecture Room A
Friday, August 5, 2005, 10:30 AM
Abstract
Determining the definite integral I of a function f over an interval
is a basic concept in mathematics. For cases that cannot be treated
analytically, quadrature rules are applied to obtain I numerically,
ideally to within a user-prescribed accuracy. Typically, f is
evaluated at stipulated points in order to implement the rule.
What can be done (a) if f is given by measurement at a number of
points, (b) if these points are arbitrarily spaced, (c) about the
measurement uncertainties? An approach is described based on
(1) regarding the measurement points as representing a curve, the
area under which is sought, (2) representing each interval between
adjacent points by a polynomial piece, (3) forming the definite
integral of each such piece, (4) summing these integrals to
approximate the required area, (5) comparing the results obtained
using polynomials of different orders, (6) propagating the
measurement uncertainties to evaluate the uncertainties associated
with the approximations.
An application to climate change is presented, viz., quantifying
the component of the Earth's Radiation Budget relating to the
incoming radiation from the sun.
Speaker Bio
Dr. Cox is a Senior Fellow at the National Physical Laboratory (NPL),
where he is a consultant mathematician. He is leader of the BIPM
Director's Advisory Group on Uncertainties, Convenor of ISO Working
Group ISO/TC 69/SC 6/WC 7 on Statistical Methods to Support
Measurement Uncertainty Evaluation, Convenor of British Standards
Panel SS/6/-/3 on Measurement Uncertainty, Convenor of British
Standards Committee SS/6 on Precision of Test Methods, lead author of
Supplement 1 to the Guide to the Expression of Uncertainty in
Measurement. Dr. Cox received a BSc in Applied Mathematics and a
PhD in numerical analysis, from City University, London, UK. His
research interests include mathematical modelling (especially applied
to measurement problems), statistical modelling (especially applied
to inter-laboratory comparisons), uncertainty evaluation, numerical
analysis, mathematical algorithms, software validation.
NIST Contact:
Charles Hagwood,
(301) 975-2846.