A symbolic algorithm for computing coefficients' accuracy in regression

https://doi.org/10.1016/B978-1-55860-247-2.50047-4Get rights and content

Abstract

Numerical law discovery and statistical regression have the same goal of summarizing data. With the aim of making them collaborate, we have integrated regression in our scientific discovery system ARC in order to discover polynomial laws. One problem to be faced was the accuracy of the coefficients determined by regression. In order to relate the coefficients of the polynomials to other parameters, ARC requires their accuracy. Since no direct method is available to compute coefficient's accuracy, we have created a heuristic method which allows to approximate them. We compare it to statistical methods we have also elaborated, and experimentation leads us to choose one or another depending on the degree of the concerned polynomial.

References (0)

Cited by (3)

View full text