Kolassa and Tanner (1999b) present an algorithm for small sample conditional confience regions for two or more parameters for any discrete regression model in the GLIM family. This method presupposes the use of approximate or exact techniques for calculating probabilities for all atoms in the sample space for some components of the vector of conditional sufficient statistics conditional on other components. Such enumeration may be performed exactly or by exact or approximate Monte Carlo, including the algorithms of Kolassa and Tanner (1994, 1999a) and Kolassa (1999). Kolassa and Tanner (1999b) present an example in which this technique is marginally more accurate than a known asymptotic technique. This paper presents an example in which all known competing asymptotic techniques fail.
展开▼
机译：Kolassa和Tanner（1999b）提出了一种算法，用于针对GLIM系列中任何离散回归模型的两个或多个参数的小样本条件置信区域。此方法以使用近似或精确技术为前提，该条件用于计算条件空间足够统计量的向量中某些分量的样本空间中所有原子的概率，该条件取决于其他分量。此类枚举可以完全或通过精确或近似的蒙特卡洛执行，包括Kolassa和Tanner（1994，1999a）和Kolassa（1999）的算法。 Kolassa and Tanner（1999b）给出了一个例子，其中该技术比已知的渐近技术略微准确。本文提供了一个示例，其中所有已知的竞争渐进技术都失败了。
展开▼
