Mathematical Sciences: Conditional Inference in the Presenceof a Nuisance Parameter

当有几个变量存在时，在进行统计推断时，通常只有其中一些是直接感兴趣的。然而，这些并不总是独立于其余部分。因此，推论通常以控制这一余数的参数的估计值为条件。根据依赖关系的性质和所使用的特定估计，主要统计推断可能非常稳定，也可能不稳定。本研究考虑了在存在多维干扰参数的条件推理中构造近似辅助统计或接近理想估计的方法。

In drawing statistical inferences when there are several variables present, often only some of these are of direct interest. However these are not always independent of the remainder. Therefore, inferences are usually drawn conditional upon the estimated values for the parameters which govern this remainder. Depending upon the nature of the dependencies and the particular estimates used, the primary statistical inferences may or may not be very stable. This research considers methods for constructing approximate ancillary statistics or close-to-ideal estimates for conditional inference in the presence of a multidimensional nuisance parameter.