Laserfiche WebLink
<br /> <br />~ 3.0 <br /> <br /> Data Analysis and Interpretation <br />' <br /> 3.1 Statistical Methods <br /> To begin to find some meaning in the data that is collected, it is necessary to use a <br /> statistical approach. This carries with it some ideas that bear discussion. It is not <br /> possible to determine with 100 percent certainty that a particular condition exists. No <br />' monitoring effort can create information that has no possibility of being wrong. The <br /> purpose of statistics is to very deliberately quantify what that possibility is. Typically <br /> this is expressed as a percentage of confidence. This value is usually dependent on the <br /> number of samples that are available. If many measurements have been taken, a <br /> statistical test may show that there is a 95 percent chance the outcome of the test is <br /> correct. In plain language this means that, for instance, if analysis shows no change in <br />' groundwater levels before and after dewatering, with a 95 percent confidence "interval", <br /> this indicates that 5 times out of every 100, this conclusion is not correct. <br />' It should be noted that the confidence interval may be as low as 80 percent, <br /> corresponding to a 1 in 5 chance of an incorrect result. It is recommended that whenever <br /> possible the confidence interval used to evaluate the results of this monitoring program <br />' be at least 90%. This leaves only a I in 10 chance of an incorrect result. <br /> There are many statistical tests and estimating procedures available to analyze data. <br /> Different tests can detect different phenomena, such as seasonality, or trends, and each <br /> test may have slightly different requirements for a meaningful result. For example, the <br /> common statistical tests, such as mean (average) and standard deviation require that a <br />' population (the group that measurements are taken from) have a normal distribution. <br /> This is the famous "bell curve" with points evenly distributed around the average. <br />' Unfortunately, many populations in the environment, such as groundwater levels, don't <br /> necessarily have a normal distribution. Special tests or procedures are needed that don't <br /> requ've the data to follow any detectable pattern. These are called non-parametric tests, <br /> and they are recommended in this case. <br /> Before any analysis is conducted, it is important to return to the question of causality. <br /> Flow in the river is known to influence the level of groundwater in the aquifer along the <br /> Cache La Poudre River. It is possible to relate flow to groundwater level and remove the <br /> influence of river flow from the data prior to analysis. Although this may not completely <br /> isolate the influence of mining on groundwater levels (local precipitation and irrigation <br /> ditch flow will also have some effect, for instance) this procedure can help correct some <br /> errors in assessing mining impacts. For this reason, flow data from the nearest USGS <br /> <br /> <br />