Browse Prior Art Database

Discrete Data Fitting Scheme

IP.com Disclosure Number: IPCOM000039443D
Original Publication Date: 1987-Jun-01
Included in the Prior Art Database: 2005-Feb-01
Document File: 2 page(s) / 13K

Publishing Venue

IBM

Related People

Gaffney, JE: AUTHOR [+2]

Abstract

A technique is disclosed for performing a linear regression analysis to find the series expansion expression which most closely approximates the logarithm of the measured data for the occurrence of errors versus time. The disclosed technique enables the user to make estimates concerning the error content of an element of software based upon measured data showing the time of occurrence of errors over a period of observation. The technique estimates the parameters of a decaying exponential curve which most closely fits the measured data. Since the rate of error discovery decreases with time, the best fit curve can be extrapolated to determine the total number of errors which can be expected to be discovered over the life of the software product.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 54% of the total text.

Page 1 of 2

Discrete Data Fitting Scheme

A technique is disclosed for performing a linear regression analysis to find the series expansion expression which most closely approximates the logarithm of the measured data for the occurrence of errors versus time. The disclosed technique enables the user to make estimates concerning the error content of an element of software based upon measured data showing the time of occurrence of errors over a period of observation. The technique estimates the parameters of a decaying exponential curve which most closely fits the measured data. Since the rate of error discovery decreases with time, the best fit curve can be extrapolated to determine the total number of errors which can be expected to be discovered over the life of the software product. Previous workers in the field have found that the rate of error discovery in the life cycle of a software product can be approximated as a decaying exponential equation as follows: r(t) = EBe-Bt (1) This earlier work can be found in [*]. Thus, the number of errors discovered over a period of time from 0 to t or N(t) is given by the integral of Equation 1 which is as follows: N(t) = E(1-e-Bt) (2) Function

N(t) may be considered to be the mean value function of a Poisson counting process. That is, the probability density function for the number of errors discovered during some interval is a Poisson distribution whose density L is a function of time. The principle of the invention disclosed herein is the performance of the estimation of the parameters of N(t), the "mean value function" of a theoretically infinite number of realizations of the Poisson error counting process, based on a single particular realization of that p...