Figure 1.

When the random grouping numbers of pools risen to three, the frequency of non-significant difference between the case and control data with normal distributions reached down to the value of that the difference calculated by individual data*. The pool numbers (i.e., numbers of pools in x-axis) was set as 2, 3, 4, and 5 and coded as G2, G3, G4, and G5; 100 times random grouping was applied to the 200 artificial data with normal distributions for the case and control populations, respectively; the means of the case are 1.0, 1.4, 1.9, 2.9 and 5.6-folds of the control and coded as D5, D4, D3, D2 and D1, respectively. *For the upper mentioned normal distribution data, calculated differences by using three random pool means of the case and control, respectively, should be the same as by using the individual data.