Journal of the Southern African Institute of Mining and Metallurgy
versão impressa ISSN 0038-223X
We consider in this paper the use of a modified version of Hotelling's statistic in the analysis of particle size distributions. The statistic can be adversely affected by the presence of outliers among the data. We propose a competitor to the statistic that is based on ranks, and hence is less sensitive to outlier effects. The results of a Monte Carlo study suggest that the rank test is highly competitive with the Hotelling test in its ability to detect differences between two mass size distributions. The calculation of the rank statistic is explained in detail and its application is illustrated on two sets of data.
Palavras-chave : mass size distributions; bias testing; rank statistic..