Laserfiche WebLink
GOF Tests to Assess Normality or Lognormality for Full Data Sets without ND Observations <br /> • Informal graphical Q-Q plot(normal probability plot) and histogram. <br /> • Shapiro-Wilk(SW)test for sample sizes less than or equal to 50. <br /> • Lilliefors test for larger sample sizes such as greater than 50. <br /> GOF Tests to Assess Normality or Lognormalityfor Left-Censored Data Sets with NDs and <br /> Multiple Detection Limits <br /> • ProUCL 4.0 can be used to perform normal GOF tests as mentioned above (for fall data) <br /> on data sets consisting of ND values. Specifically, normal or lognormal GOF tests can be <br /> performed using detected data values only. <br /> • The normal or lognormal GOF tests can also be performed on data sets (detected values <br /> and extrapolated NDs) obtained using one of the regression on order statistics (ROS) <br /> methods. The details of constructing Q-Q plots (normal and lognormal) and performing <br /> ROS on data sets with multiple DLs are given in ProUCL 4.0 Technical Guide. <br /> • The three ROS methods available in ProUCL 4.0 are the normal ROS, lognormal ROS <br /> (also known as robust ROS), and Gamma ROS methods. <br /> • ProUCL 4.0 can be used to generate additional columns (with suitable headings assigned <br /> by ProUCL 4.0) of data consisting of the detected data and extrapolated nondetect data. <br /> Goodness-of-Fit Test for Gamma Distribution <br /> Gamma GOF Tests for Full Data Sets without ND Observations <br /> • Informal graphical quantile-quantile (Q-Q)plot(gamma probability plot) and histogram. <br /> • Kolmogorov-Smirnov test for sample sizes in the range 4-2500 (critical values computed <br /> using Monte Carlo simulations) and values of the estimated shape parameter,k, in the <br /> interval [0.01, 100.01. <br /> • Anderson-Darling test for sample sizes in the range 4-2500 (critical values computed <br /> using Monte Carlo simulations) and values of the estimated shape parameter, k, in the <br /> interval [0.01, 100.0]. <br /> Gamma GOF Tests for Left-Censored Data Sets with NDs and Multiple Detection Limit <br /> • ProUCL 4.0 can be used to perform gamma GOF tests on data sets consisting of ND <br /> values. Specifically, gamma GOF tests can be performed on data set consisting of only <br /> detected data. <br /> • The gamma GOF tests listed above can also be used on data sets (detected values and <br /> extrapolated NDs) obtained using one of the regression on order statistics (ROS)method <br /> as incorporated in ProUCL 4.0. The details of constructing gamma Q-Q plots and <br /> performing ROS on data sets with multiple detection limits are given in ProUCL 4.0 <br /> Technical Guide. <br />