Avaliação de um programa interlaboratorial de controle de qualidade de resultados de análise de solo

Abstract

The monitoring of the quality of soil analysis reduces errors in the recommendations of liming and fertilizers. The program of quality control of ROLAS-RS/SC network evaluates the analysis accuracy by testing four soil samples monthly for a year. However, the data sets must have normal distribution and absence of outliers to insure that the median can be considered as an estimation of the true value from the soil samples. Therefore, the objectives of this work were to check the normal distribution, identify outliers and evaluate procedures of accuracy calculations, such as how these aspects may affect laboratory accuracy. The Lilliefors test was run to check the normality and the outliers were identified through the quartile test. The outliers were eliminated from data population. The substitution of the median by the average as criterion of central reference was tested as well as the procedure for calculating the accuracy per attribute instead of annual average accuracy. Only 48% of the data followed normal distribution. The exclusion of outliers improved the analyzes with normal distribution up to 65%, which decreased laboratories with minimal accuracy required by the network. When data sets have normal distribution, the average shows a better estimation than the median and the procedure of calculating the annual average accuracy may hide attributes with inferior accuracy than the minimal average required.