City Research Online

Abstract

Multiple linear regression is one of the most widely used predictive models across a broad range of applications. We propose a novel regression framework that, instead of minimising the aggregate prediction error in the dependent variable, distributes the total prediction error evenly across all parameters. This approach is particularly suitable for data affected by substantial noise, as is often the case for time series data where structural changes and evolving trends are common. We provide a theoretical characterisation of our proposed estimator, named Parity Regression. Its properties are compared with those of existing penalised and shrinkage estimators in the literature. Both synthetic experiments and real-data applications demonstrate that the theoretical guarantees of the proposed method are reflected in practice.

Publication Type:	Other (Preprint)
Publisher Keywords:	Ordinary Least Square, Parity, Ridge Regression, Shrinkage Estimation
Subjects:	H Social Sciences > HG Finance
Departments:	Bayes Business School
SWORD Depositor:	Symplectic Administrator

Export

Downloads