作者: Bruce G. Lindsay , Kathryn Roeder
DOI: 10.1080/01621459.1987.10478496
关键词:
摘要: Abstract It is shown that there an analog to the score function (the derivative of log-likelihood with respect parameter), here called difference score, which plays important theoretical role in integer parameter models. A unified treatment models can be obtained by recognizing many commonly used have a linear (LDS). established exact results available for exponential family also hold LDS In particular we find information lower bound, show uniqueness maximum likelihood estimator, establish monotone ratio, and optimality properties two-sided tests. if one has multiparameter conditional depending on N alone, then these apply as well. when why usual methodology will class nonstandard natural approxima...