Laserfiche WebLink
APPENDIX C <br /> TREND ANALYSIS ETH®D <br /> The Kendall Rank Correlation test for trend, also called the Mann-Kendall test, is a <br /> nonparametric,rank correlation method which can be used for evaluating correlations between <br /> time and constituent concentrations. The test is described in Statistical Methods for <br /> Environmental Pollution Monitoring(Richard 0.Gilbert,Van Norstrand Company,New York, <br /> 1987) and in Nonparametric Statistical Methods (Myles Hollander and Douglas A. Wolfe, <br /> John Wiley and Sons, New York, 1973). Its use is recommended in An Evaluation of Trend <br /> Detection Techniques for Use in Water Quality Monitoring Programs (J.C. Loftis et al., <br /> Environmental Protection Agency, 1989). <br /> The Kendall Rank Correlation test is nonparametric, meaning that it does not depend on an <br /> assumption of a particular underlying distribution. The test uses only the relative magnitudes <br /> of the data rather than their actual values; therefore, missing values are allowed, and <br /> constituents which are not detected can still be used in the statistical analysis by assigning <br /> values equal to half their detection limits. <br /> To determine if constituents exhibit statistically significant trends, the Kendall Rank <br /> Correlation Test for trend is applied using the StatView computer program (Abacus Concepts, <br /> 1992). The null hypothesis, Ho, is that no trend of constituent concentrations exists over time. <br /> The alternative hypothesis, Ha, is that either an upward or downward trend of constituent <br /> concentrations exists. <br /> SAGpjM3M390011.8a1-94 1fc:9 C-1 Rev.0,October 28,1994 <br />