Interval Regression Model Using L(p,q) -Estimator

Abstract

In this paper we introduce an interval regression model, having interval coefficients and crisp input data, and propose a L(p,q)-estimator for the interval coefficients. The L(p,q)-estimator uses one Lp-norm to estimate the left end point of the interval coefficients and independently another Lq-norm for the width of the intervals. An example is presented showing how the results of the regression model vary as we change p and q which implies we need to find the “best" p and q. An error measure is defined and we use response surface methodology to search for the optimal values for p and q to minimize this error.