On the Analysis of Grouped Extreme Value Data with GLIM of Fixed Source
研究了如何用广义线性模型直接估计分组极值数据中极值分布的位置参数，将多项分布似然转化为独立二项分布似然以便处理，并以微芯片设计为例说明方法。
A problem is studied in which the parameters of a linear model for the location of the extreme value distribution (EVD) are to be estimated. Observations on this EVD are indirect and the data consist of counts in categories with known boundaries, so that the data are grouped EV data. To estimate the model parameters in a direct way with GLIM, the likelihood of the (K−1)-dimensional multinomial distribution for the data is re-expressed as the likelihood of K−1 independent binomials. These can easily be handled with the GLIM system. The method is illustrated with an example on microchip design