摘要
When the statistics distribution of interested is complex or vague, Bootstrap method may be used to estimate standard error and confidence interval. However, the results derived from parametric and nonparametric Bootstrap methods applied to hierarchically structured data may be different. According to the comparing, we found and concluded that nonparametric Bootstrap method is relating seldom affected by the distribution in models. In terms of the nonparametric Bootstrap sampling, the effectiveness of sampling on the highest level unit is more satisfactory than lower level units. The estimation of confidence intervals and hypothesis testing of intraclass corelation coefficients is discussed based on multilevel generalized linear models.
When the statistics distribution of interested is complex or vague, Bootstrap method may be used to estimate standard error and confidence interval. However, the results derived from parametric and nonparametric Bootstrap methods applied to hierarchically structured data may be different. According to the comparing, we found and concluded that nonparametric Bootstrap method is relating seldom affected by the distribution in models. In terms of the nonparametric Bootstrap sampling, the effectiveness of sampling on the highest level unit is more satisfactory than lower level units. The estimation of confidence intervals and hypothesis testing of intraclass corelation coefficients is discussed based on multilevel generalized linear models.
出处
《统计研究》
CSSCI
北大核心
2002年第3期55-59,共5页
Statistical Research
