当{xi}有界及μ满足一定条件时,二值响应变量拟似然方程的解为真参数β0的相合估计.人们努力想把结果推广到{xi}无界的情形,迄今并未取得多少有意义的结果.本文通过构造正反3个例子指出:这个结论不可能以任何有一定普遍意义的形式推广到无界的情形.
Abstract: If xi is bounded and μ satisfies some conditions, the solution of two-values response variable of quasi-likelihood equation ^n∑i=1xi(yi-μ(x'iβ)=0 is a consistent estimate of the true parameter β0. Many researcher tried to extended to unbounded condition, but did not obtain any significant result. Here we show that the conclusion cannot be extended to unbounded xi in any form with some generality.