A Regression Formulation of the Matrix Estimation Problem

Publisher Website
Google Scholar
Abstract
Matrices are widely used in transportation planning to represent the distribution of characteristics or as origin-destination matrices. Developing such matrices by means of surveys is expensive and time consuming, and once the survey data are collected and compiled the matrices are rapidly outdated. Other methods which are commonly used are unable to include all available data or to provide a measure of the uncertainty of the estimates. This paper formulates a quadratic programming method to estimate matrix entry estimates as an equivalent constrained generalized least squares estimation problem. As well as being able to include any available information in the form of constraints, the variance-covariance matrix of the entry estimates may be found and confidence intervals calculated for matrix entry estimates with some added distributional assumptions. The problem of updating the proportions of nationwide automobile trips by purpose and trip length from 1970 to 1977 is included as a simple example to illustrate the method.

This publication has 0 references indexed in Scilit: