Backtesting forecast accuracy

TL;DR

This paper introduces a geometric mean-based statistical test to evaluate the accuracy of predictive models for positive continuous variables, with validation through simulations and an insurance application.

Contribution

It proposes a novel geometric mean-based test for forecast accuracy and compares its performance with existing methods.

Findings

The test effectively distinguishes accurate from inaccurate models.

Simulation results show competitive power against alternative procedures.

Application to insurance claims demonstrates practical utility.

Abstract

A statistical test based on the geometric mean is proposed to determine if a predictive model should be rejected or not, when the quantity of interest is a strictly positive continuous random variable. A simulation study is performed to compare test power performance against an alternative procedure, and an application to insurance claims reserving is illustrated.