|
contaminant относится как раз к population Contaminant populations:
In contradistinction to the majority distributions, four of the six contaminating population models considered differ in how the xi are distributed but have the same u(xi) distribution, u(xi); that is, they are uniformly distributed random numbers from 0.1 to 4. In addition, two contaminant models examine the relative importance of under- and over-estimation of u(xi) within these minimum and maximum limits. In addition, two contaminant models examine the relative importance of under- and over-estimation of u(xi) within these minimum and maximum limits. The small but non-zero lower limit and the modest magnitude of the upper limit keep the uncertainties from being ‘‘unrealistically’’ small or too huge relative to those of the majority populations. Figure 4 displays dot-and-bar graphs of 20 {xi, u(xi)} exemplar data for each of the six
contamination models, along with the PDFs for the exemplar data and the N(0,1) reference. Figure 5 displays the joint distributions for 4,000 data for the six contaminant populations considered.
In contaminant population E1 the xi * UR[-8,8], making about half of the contaminants readily identifiable value outliers. In contrast, there are no
value outliers at all in the xi * N(0,1) of contaminant population E2 (Figs. 4, 5, upper right). In population E3 (Figs. 4, 5, middle left) the xi * UR[-4,4], giving a modest proportion of marginally identifiable value outliers. The xi in these three contaminant populations are distributed symmetrically about zero, ensuring that the true location of the mixed population is identical to that of the
majority.
|