论文部分内容阅读
目的 探讨四格表资料 χ2 检验是否需要连续性校正以及是否需要用确切概率法。方法 用计算机模拟从两个总体 (π1=π2 )中随机抽取 10 0 0 0 0个样本 ,用四格表专用公式、校正公式和确切概率法进行假设检验 ,统计三种方法犯第一类错误的频率 ;模拟从两个总体 (π1≠π2 )中随机抽取 10 0 0 0 0个样本 ,用以上三种方法进行假设检验 ,统计判断正确的频率 (检验效能 )。结果 专用公式犯第一类错误的频率比校正公式和确切概率法更接近检验水准α ;专用公式的检验效能高于校正公式和确切概率法。结论 对四格表资料不宜进行连续性校正 ,确切概率法也未必是最好的方法
Objective To investigate whether the two-table data χ2 test requires continuity correction and whether exact probability method is needed. Methods A computer simulation was used to randomly sample 100 000 samples from two populations (π1=π2 ). Hypothesis testing was performed using a four-grid table-specific formula, a correction formula, and an exact probability method. Three methods were used to make the first error. The frequency; the simulation randomly selected 10 0 0 0 samples from two populations (π 1 ≠ π 2 ), using the above three methods for hypothesis testing, statistically determine the correct frequency (test effectiveness). Results The frequency of the first type error of the special formula is closer to the test level α than the correction formula and the exact probability method; the test efficiency of the special formula is higher than the correction formula and the exact probability method. Conclusion It is not advisable to carry out continuous correction on the data of the four-table. The exact probability method may not be the best method.