Computes the location estimator based on van der Waerden scores (sometimes also referred to as normal scores).
vdw.loc(x, int.diff = 10, maxiter = 1000, na.action = na.fail)the van der Waerden score estimator of location.
a numeric vector.
no longer used, kept for compatibility with earlier versions.
no longer used, kept for compatibility with earlier versions.
a function which indicates what should happen when the data contain 'NA's. Default is to fail.
Klaus Nordhausen
The estimator is the Hodges-Lehmann type estimator based on the signed-rank statistic with van der Waerden scores $$T(\mu) = \sum_{i=1}^n sign(x_i - \mu) \Phi^{-1}\left(\frac{1}{2} + \frac{R_i^+(\mu)}{2(n+1)}\right),$$ where \(R_i^+(\mu)\) is the rank of \(|x_i - \mu|\) among \(|x_1 - \mu|, \ldots, |x_n - \mu|\). The function \(T\) is nonincreasing and constant between neighbouring Walsh averages \((x_i + x_j)/2\), \(i \le j\), and the estimate is $$\hat\mu = \frac{1}{2}\left(\sup\{\mu : T(\mu) > 0\} + \inf\{\mu : T(\mu) < 0\}\right).$$ It is computed exactly by a bisection over the sorted Walsh averages.
With Wilcoxon scores instead of van der Waerden scores the same definition gives the
Hodges-Lehmann estimator hl.loc.
Up to version 1.1-3 the estimate was obtained by a linear interpolation between two Walsh averages, which gave only an approximation of the estimate and could fail.
Hodges, J.L. and Lehmann, E.L. (1963), Estimates of location based on rank tests, The Annals of Mathematical Statistics, 34, 598--611.
Hettmansperger, T.P. and McKean, J.W. (1998), Robust Nonparametric Statistical Methods, London, Arnold.
hl.loc
set.seed(1)
x <- rt(100, df = 3)
vdw.loc(x)
rm(.Random.seed)
Run the code above in your browser using DataLab