Laserfiche WebLink
Q138 <br />the trend is less significant because there is less information in serially <br />correlated data. The difference between the unadjusted and adjusted signifi- <br />cance levels increases as the strength of-the serial correlation increases. <br />The trend slope listed in tables 13 through 21 is the seasonal Kendall <br />slope estimator, as defined by Hirsch and others (1982). A negative trend <br />slope indicates a decrease in the property or concentration of the constituent <br />with time; a positive trend slope indicates an increase in the property or <br />concentration of the constituent with time. The seasonal Kendall slope <br />estimator, the trend slope, is the median of all possible differences in the <br />time - series data within the same month and provides an estimate of the median <br />annual change in the data. <br />In instances where the distribution of the data was highly skewed (the <br />data distribution is asymmetrical; the mean is different from the median), a <br />trend is unlikely to be linear; therefore, the trend -slope estimator computed <br />from actual data values is not appropriate. Transforming the data to <br />logarithms will linearize the trend if the annual changes are proportional to <br />each other. The significance of the trend, which is based on the ranks of the <br />data, is not affected by this transformation. Most of the trend analyses in <br />this study were made on the actual data values; however, the logarithms of the <br />number of colonies per 100 mL was used for trend analyses of all bacterial <br />data. The trend slope for the bacterial data in tables 13 through 21 is the <br />multiplicative change in median value each year because of the logarithm <br />conversion. The change is not constant with time but increases or decreases <br />with time. The change in median has been detransformed and is the overall <br />change as a percentage of the estimated median for the first year of record in <br />the original data units. <br />Significant trend slopes are identified in tables 13 through 21 by using <br />the following criteria: <br />1. Moderately significant ( *), if the significance level is less than <br />or equal to 0.1 and greater than 0.05. <br />2. Significant ( **), if the significance level is less than or equal <br />to 0.05 and greater than 0.01. <br />3. Very significant ( %'^"), if the significance level is less than or <br />equal to 0.01. <br />If the significance level (p- value) is greater than 0.1, there is a greater <br />than 10- percent chance that there is no real trend. Selection of the signifi- <br />cance level used in the above identification, if more than one significance <br />level was reported for a property or constituent, was based on the order: <br />(1) Significance level adjusted for serial correlation; and (2) unadjusted <br />significance level, if there was no adjusted significance level reported. <br />For example, in table 13, the trend slope for water temperature is not <br />significant. In table 13, the trend slope for pH was determined to be <br />significant because the adjusted significance level (greater than 10 years of <br />record) 0.0497 is less than or equal to 0.05 and greater than 0.01. Also in <br />table 13, the trend slope for fecal streptococcus is very significant because <br />the unadjusted significance level is 0.0100. <br />15 <br />