Laserfiche WebLink
quantile-quantile (Q-Q)plots, box plots, and histograms. The graphical displays of Q-Q plot and <br /> box plot are also useful to visually identify outliers that may be present in a data set. It is noted <br /> that,the classical test statistics such as the Dixon test and the Rosner test get distorted by the <br /> presence of the same outlying observations that those tests are supposed to identify. Therefore, <br /> those test statistic (Dixon and Rosner)results should always be supplemented by the graphical <br /> displays to confirm the presence of outliers (and potential multiple populations) in a data set. <br /> Alternately, the use of robust and resistant outlier identification methods (Singh and Nocerino, <br /> 1995) is recommended to identify outliers. The robust outlier identification methods are beyond <br /> the scope of ProUCL 4.0. <br /> The proper disposition of outliers to include or not to include outliers in the computation of <br /> various statistics should be determined by the project team, site experts, and the decision makers <br /> involved in the project. In an effort to determine the influence of outliers on the statistics of <br /> interest, it is suggested to compute the various statistics based upon data sets with and without <br /> the outliers. This extra step should help the project team in determining the proper disposition of <br /> outliers. These issues have also been discussed in detail in ProUCL 4.0 Technical Guide. <br /> Screen Shots Generated By ProUCL 4.0 <br /> s Box Plats for Mn(1), PAn(E). Mn(9) — -- <br /> m o <br /> OJbO .i3X <br /> A <br /> p"WO <br /> a <br /> zatuno <br /> O ivaeao _ s iEffiJ <br /> ,.a ... �o- <br /> ,Iwo '.. <br />