A Constrained Spatial Autoregressive Model for Interval-valued data

TL;DR

This paper introduces a new spatial autoregressive model for interval-valued data that accounts for spatial dependencies and incorporates inequality constraints to enhance prediction accuracy.

Contribution

It develops a novel linear model for spatially correlated interval data, integrating inequality constraints and a combined grid search and constrained least squares algorithm.

Findings

The model effectively captures spatial dependencies in interval data.

Numerical experiments show improved prediction accuracy.

Application to weather data demonstrates practical usefulness.

Abstract

Interval-valued data receives much attention due to its wide applications in the fields of finance, econometrics, meteorology and medicine. However, most regression models developed for interval-valued data assume observations are mutually independent, not adapted to the scenario that individuals are spatially correlated. We propose a new linear model to accommodate to areal-type spatial dependency existed in interval-valued data. Specifically, spatial correlation among centers of responses are considered. To improve the new model's prediction accuracy, we add three inequality constrains. Parameters are obtained by an algorithm combining grid search technique and the constrained least squares method. Numerical experiments are designed to examine prediction performances of the proposed model. We also employ a weather dataset to demonstrate usefulness of our model.