Algoritmo para Suavizaςão de Dados Via Ajuste de Curvas por Mínimos Quadrados com Detecςão de Dados Inválidos

Abstract

An algorithm for data smoothing or validation by least square curve fitting, with a Chi-Square procedure of curve degree automatic selection is presented. The algorithm comprises an invalid data detection procedure based on the F distribution, which reaches its objective monitoring the contribution of each data point in the
quadratic residuum sum. The procedure application allows handling with a ratio of invalid to valid data greater than the usual procedure of data editing based on a tolerance zone around of adjusted curve. This can be clearly observed by comparing the tests results.