Abstract

The Interpretable Components (IC) use restrictions in order to have a better interpretation of the coefficients related to a Principal Component (PC). The efficiency of a (IC) due to a (PC) is made in relation to the angle formed between the Components, it being desirable the lowest value. In this context, an alternative to enrich this validation is the use of measures and circular distances so far not applied for this purpose. Given this motivation, this paper aims to propose the use of these measures to evaluate the robustness of the (IC) compared to samples contaminated with outliers, using measures and circular distances. According to various scenarios evaluated through Monte Carlo simulation, it was concluded that the use of these measures are recommended to validate the (PC) and when considering the Toeplitz correlation structure, IC were more robust in relation to the presence of outliers.

摘要

可解释组件 (IC) 使用限制，以便更好地解释与主组件 (PC) 相关的系数。由于 (PC) 而导致的 (IC) 的效率与组件之间形成的角度有关，希望是最低值。在这种情况下，丰富此验证的另一种方法是使用迄今为止尚未用于此目的的度量和圆形距离。鉴于这种动机，本文旨在提出使用这些措施来评估 (IC) 与被异常值污染的样本相比的稳健性，使用措施和圆形距离。根据通过蒙特卡罗模拟评估的各种场景，得出的结论是，建议使用这些措施来验证 (PC)，并且在考虑 Toeplitz 相关结构时，