On a non-parametric confidence interval for the regression slope

TL;DR

This paper critically examines a non-parametric confidence interval for the regression slope derived from Tukey's methodology applied to Theil's regression, revealing potential issues with confidence level accuracy and emphasizing the importance of assumption validation.

Contribution

It provides theoretical proofs for the confidence interval's properties with small data samples and warns against unverified method combinations in statistical practice.

Findings

Theil's regression confidence interval can significantly underperform in maintaining true confidence levels.

Theoretical validation is provided for small sample cases (4 and 5 data points).

Practical example demonstrates the importance of assumption checking in statistical methods.

Abstract

We investigate an application of the Tukey's methodology in Theil's regression to obtain a confidence interval for the true slope in the straight line regression model with not necessarily normal errors. This specific approach is implemented since 2005 in a package of the software R; however, without any theoretical background. We illustrate by Monte Carlo simulations, that this methodology, unlike the classical Theil's approach based on Kendall's tau, seriously deflates the true confidence level of the resulting interval. We provide also rigorous proofs in case of four data points (in general) and in case of five data points (under some additional conditions); together with a real life methods usage example in the latter case. Summing up, we demonstrate that one should never combine statistical methods without checking the assumptions of their usage and we also give a warning to the already wide community of R users of Theil's regression from various fields of science.