Linear regression with known noise distribution up to a scale: the reward of not using the OLSE

Current Robotics Reports · Springer Nature · 2026-09-02 · DOI: 10.1007/s10260-026-00868-7

Аннотация

While the ordinary least squares estimator (OLSE) is still the most widely used estimator in linear regression models, other estimators can be more efficient when the error distribution is not Gaussian. In this paper, our goal is to evaluate this efficiency in the case of the Maximum Likelihood estimator (MLE) when the noise distribution belongs to a scale family. Under some regularity conditions, we show that $$(\widehat{\beta }_n, \widehat{s}_n)$$, the MLE of the unknown regression vector $$\beta _0$$ and the scale parameter $$s_0$$, exists and derive the expression of the asymptotic efficiency of $$\widehat{\beta }_n$$ over the OLSE. For three given scale families of densities, we quantify the true statistical gain of the MLE as a function of their deviation from the Gaussian family. To illustrate the theory, we present simulation results for different settings and also compare the MLE to the OLSE for the real market fish dataset.

Читать у издателя →

Похожие статьи