Aplicação da Estatística Direcional no Controle de Qualidade Cartográfica

Application of Directional Statistics in Cartographic Quality Control

Abstract

A cartographic product is considered positionally accurate when it is precise and not biased. Trend analysis is generally performed using the Student's t test, which requires that the sample follows a normal distribution. One solution is the use of circular descriptive statistics based on Directional Mean and Circular Variance, which do not assume normality. However, analyzes of circular descriptive statistics may be inadequate and due care is not taken. Therefore, it is necessary to add more robust processes on directional statistics for positional assessment, such as joint analyzes of the main descriptive parameters, application of statistical tests of normality and directional uniformity. The use of these processes is evaluated in this work with the aim of identifying solutions to avoid failures in trend detection in cartographic data. These failures were exemplified by simulated models, which demonstrated situations in which the analysis of Directional Mean and Circular Variance may not be effective. It was also proposed to use statistical tests, joint analysis of several descriptive parameters and graphs of directional statistics as a complement to detect trends. This procedure was applied to two Digital Surface Models (MDS), with trends detected in both.