STATISTICAL PROCEDURES FOR CORROSION STUDIES

Decision-makers evaluating corrosion treatment alternatives under the Lead and Copper Rule need the statistical tools to choose the optimum treatment from pipe loop studies. In many cases, the data generated by pipe loop studies do not follow a normal distribution; therefore, non parametric statistics apply. Techniques that are discussed include det ermination of data normality using the Kolmogorov-Smirnov and chi-squa re tests, determination of stabilization using the Spearman coefficien t, and comparison of treatments using the Wilcoxon signed ranks or ran k sum test. Other data issues that are important to evaluating corrosi on studies include determining sample size and frequency, determining the confidence and accuracy of results, and evaluating data outliers.