Figure 2. Factors controlling the departure of real and nominal weights.

The red curves correspond to a prevalence ratio of 10, and the black curves correspond to a prevalence ratio of 2. For each fixed prevalence ratio, two sample sizes (plotting symbol  = 1) and (plotting symbol  = 2) are shown. For a fixed prevalence ratio and sample size, one can vary the number of positives in class 0 (C₀), and compute the corresponding number of positives in class 1 (C₁). The x-axis is C₀. One can then compute the nominal and real weights from C₀, C₁, N₀, and N₁. The primary determinate for departures between the real and nominal weights appears to be the number of positives in the sample (x-axis), and not the total sample size (1 versus 2 plotting symbol). The apparent prevalence ratio (red versus black) appears to play a minor secondary role.