基于蒙特卡洛模拟的四格表独立性检验研究

摘要

文章通过给出四格表三种独立性检验统计量的蒙特卡洛模拟算法,研究不同样本量和参数条件下各检验统计量犯第一类错误的概率和功效,并进行统计量优良性对比.结果表明,在样本量较小时,似然比检验优于卡方检验和优比检验;而随着样本量的增加,三种检验统计量都能较为有效地控制犯第一类错误的概率,并具有较高的检验功效.在此基础上,将上述三种检验统计量应用于实际案例的数据分析中.%In this paper, a Monte Carlo simulation algorithm is presented for three kinds of independence test statistics of fourfold table.And then, in order to test the optimality of the statistics, it uses the algorithm to obtain the Type I error probability and power of each statistics in different sample sizes and parameters.The result shows that the likelihood ratio test is better than the other methods when the sample size is small.Moreover, with the increase of the sample size, the three test statistics can more effectively control the probability of Type I error, and have a high power of the test.On this basis, the above-mentioned three test statistics are applied to the data analysis of the real cases.