As an illustration, two sets of weights are given in Table 1, and percentiles are calculated for each set of weights as well as for the simple median (equal weights). As the first set of weights are symmetric, the weighted median in this case equals the simple median. However, less weight is given to outlying estimates, and the empirical distribution function (Fig. 3, red line) is close to the median value across a wider range of the distribution. Confidence intervals for the weighted median, which can be obtained by a parametric bootstrap method, should therefore be narrower. In the second set of weights, smaller estimates happen to receive more weight (Fig. 3, blue line). The weighted median estimate will be interpolated between ratio estimates β^3 and β^4, but will be closer to β^4 as the percentile p 4 is closest to 50%. The exact weighted median estimate in this case will be β^WM=β^3+(β^4−β^3)×50−27.7852.78−27.78.The weighted median can also be thought of as the simple median from a set of values (a pseudopopulation) in which the ratio estimate β^1 for variant 1
