Distributions minimizing Fisher information for scale in \(\varepsilon\)- contamination neighbourhoods (Q1199012)

This paper deals with the problem of finding those distributions minimizing Fisher information for scale in general \(\varepsilon\)- contamination neighbourhood models \[ {\mathcal G}_ \varepsilon(G)=\{F:F=(1- \varepsilon)G+\varepsilon H;\;\varepsilon \text{ fixed},\;H\text{ arbitrary}\} \] of distribution functions \(G\) satisfying certain mild conditions; \(G\) is not necessarily symmetric. The general form of the solution given here is a generalization of a result of \textit{P. J. Huber} [Robust statistics. (1981; Zbl 0536.62025)] in the case of \(G\) being normal.