Laserfiche WebLink
<br />Method of Analvsis <br />A pair of statisticians was hired to determine and carry out the statistical methods and <br />tests that would be most appropriate and meaningful for the purpose of this analysis. <br />Several key factors were tal<en into consideration when performing the analysis. The <br />change in actual ET rates from year to year can vary by several inches. Therefore, if one <br />year was extremely dry, households would be much more lil<ely to use more water in that <br />year as compared with a year that was wetter than normal. Without tal<ing ET rates into <br />account, changes in water usage could be incorrectly attributed to Slow the Flow <br />Colorado instead of the simple change in ET requirements. Additionally, the statisticians <br />pointed out that without comparing water usage to landscape size, water consumption <br />data is not meaningful because large water users may be using an appropriate amount of <br />water given their landscape size and water requirements. <br /> <br />Although performing an analysis that expresses the results in terms of gallons saved <br />would be desirable, the results of such an analysis would not be statistically meaningful. <br />A large variance in the number of gallons used by each household existed in the water <br />records. The watering habits of households that were using extremely large amounts of <br />water would overshadow any changes made by households using less amounts of water. <br />Therefore, conveying results in terms of gallons saved would not be an effective way of <br />analyzing the Slow the Flow Colorado Program. In order to conduct a statistically <br />meaningful analysis of water consumption, data would need to be compared against a <br />mean. <br /> <br />It was determined that percent above or below (noted as +/- ) ET would be the most <br />accurate and effective means of comparison. Percent +/- ET takes into account the ET <br />rate for the year being analyzed as well as the amount of water used in relation to the <br />household's landscape size. Moreover, since all of the data was expressed as a ratio, the <br />problem of large water users overshadowing the rest of the group was considerably <br />reduced. <br /> <br />The study groups were analyzed in several different configurations for the purposes of <br />performing statistical tests. Each group was analyzed as a whole for the pre-inspection <br />year and for the post-inspection year. Each group was also brol<en up into two sub- <br />groups, which included households that watered above ET prior to the inspection and <br />households that watered below ET prior to the inspection. This was done to observe <br />whether these two groups behaved differently following the inspection since pre- <br />inspection behaviors were initially different. <br /> <br />The standard deviation was calculated for each of the abovementioned groups and was <br />used to describe the distribution of the data and to form a test statistic (such as at-test). <br />A normality test was also performed for each of these groups to test whether or not the <br />data was normally distributed. The results of the normality tests helped dictate which <br />statistical methods were most appropriate for the analysis. <br /> <br />8 <br />