p值的使用准则及其蒙特卡罗估计

Use criteria of p value and its monte carlo estimation

利用蒙特卡罗方法直观探讨了p值的含义与使用准则,在检验统计量难以得到精确分布时,通过生成参考数据去逼近目标统计量的分布,进而模拟假设检验的p值,最后给出了卡方拟合优度检验p值的蒙特卡罗估计,得到目标统计量的渐近分布与真实分布在样本量较少时误差大,建议使用蒙特卡罗检验.

Using Monte Carlo method,the meaning and use criteria of p-value are discussed intuitively.When it is difficult to obtain accurate distribution of test statistics,the distribution of target statistics is approximated by generating reference data,and then the p-value of hypothesis test is simulated.Finally,the Monte Carlo estimation of p-value of chi-square goodness of fit test is given.It is obtained that the asymptotic distribution of target statistics is compared with the real distribution in sample size.Monte Carlo test is recommended for small time error.