摘要
目的通过计算机模拟抽样,对病例对照研究资料logistic回归模型参数极大似然估计方法的应用条件进行探讨,为正确应用该法提供理论基础和应用指导。方法利用SAS函数抽取四格表资料,应用LogXact软件计算MLE及精确推断结果,对其进行比较。结果当暴露比低于0.1,尤其是OR比较大的情况下,MLE法必须满足大样本条件;随着暴露比的增加,MLE法对样本含量的要求可以放宽,暴露比为0.2,每组样本例数可为15例。结论MLE法估计lo-gistic模型参数对样本含量的要求可以有条件的放宽;精确推断是MLE法的补充。
Objective Using computer simulation technique, to discuss the application condition of maximum likelihood estimated method （MLE） in logistic regression for case - control study. Methods Using functions of SAS software, to obtain 2 × 2 contingency tables. After comparing the results of maximum likelihood estimate and exact inference for each table by using LogXact software, we obtain the basic condition of application on MLE in case - control studies. Results When exposed ratio under 0.1, especially relative risk of disease is large, MLE requests large sample data. This request can be relaxed along with the increase of exposed ratio. When exposed ratio was 0.2, sample size of each group can be 15 cases. Conclusion The request for sample size can be relaxed with special condition when we use MLE method to estimate the parameters of logistic regression model. Exact inference is supplement of MLE.
出处
《中国卫生统计》
CSCD
北大核心
2006年第3期206-208,共3页
Chinese Journal of Health Statistics
