Robust WV CI

[stephaneguerrier]

The CI for the WV are off. If the efficiency is 60% the robust CI at the 95% confidence level should be about 30% (100_( diff(c(-2,2)_sqrt(1/0.6))/diff(c(-2,2)) - 1)) wider then the classical. In our code, it is about 2 times wider for the first scale... Example:

x = rnorm(1000) wv1 = wvar(x) wv2 = wvar(x, robust = T) (wv2$ci_high - wv2$ci_low)/(wv1$ci_high - wv1$ci_low) [1] 2.019610 1.788926 1.540864 1.394738 1.257274 1.232275 1.156168 1.320823 1.350617