A highly efficient L-estimator for the location parameter of the Cauchy distribution (Q626257)

Let \(X_i\), \(i=1,\dots,n\), be an i.i.d. sample from the Cauchy distribution with unknown scale and location parameters. The author considers an estimate \(T_n\) for the location parameter of the form \(T_n=\sum_{i=1}^m w_i X_{(i)}\), where \(X_{(i)}\) is the \(i\)-th order statistics of \((X_i)_{i=1}^n\), \[ w_i=C^{-1}\cos^{2+\varepsilon_n}\pi \left({i-0.5\over n}-0.5\right) \cos 2\pi\left({i-0.5\over n}-0.5\right), \] where \(C\) is a normalizing constant determined by \(\sum w_i=1\). For \(\varepsilon_n=0\) this estimate coincides with Chernoff's asymptotically efficient estimate for \(\vartheta\). The sequence \(\varepsilon_n\) is introduced to improve the finite sample performance of the estimate. Its form \(\varepsilon_n=8/n^{1/2}+460/n^2\) is obtained via regression fitting of the simulation results.