Power robustification of approximately linear tests

摘要: Abstract We present a general method of improving the power linear and approximately tests when deviations from translation family distributions must be taken into account. This involves combination statistic measuring location quadratic change shape underlying distribution. The (“funnel tests”) are constructed as certain Bayes tests. In they gain sizeable amount over adapted to distribution occurs, while losing little for alternatives (“power robustification”). introduce concept funnel in an Gaussian framework first. effect robustification is studied by means function expansion, which applies large class sharing invariance property. characterized maximin property region defined rotational cone. idea const...