中文摘要：

目的通过对不同时间点间存在相关、无残留效应的N-of-1定量数据进行模拟研究，比较不同检验方法的统计性能。方法模拟参数设样本量为10（模型1），研究周期为3（模型2-4），不同时间点问相关系数为0．8（模型5-7），无残留效应，根据固定的效应差值产生多元正态分布数据，建立配对t检验、混合效应模型和差值的混合效应模型。使用效应差值估计值的Ⅰ类错误、检验功效、平均误差（ME），平均绝对误差（AE），均方误差（RMSE）评价各种模型。结果所有模型估计值的均数都非常接近效应差值，所有模型估计值的ME、AE、RMSE都较小。除了模型7，其他模型的I类错误概率都约等于0．05。随着效应差值的增大，所有模型的检验功效都随之增大。当两组的效应差值较小时（〈1．0），模型5的检验功效最大，模型2至模型4的功效较小。当两组的效应差值较大时（≥1．0），所有模型的功效都小于0．010。结论混合效应模型比配对t检验更适合存在相关关系的N-of-1数据。混合效应模型的效果优于差值的混合效应模型，效果最优的模型是CS结构的混合效应模型。

英文摘要：

Objective To compare the different models by u- sing simulation study based on multivariate normal distribution data of N- of-1 study, which had correlation, no carry-over effect among different points. Methods Assume that sample size was 10, the cycle was 3, corre- lation coefficient between different point was 0. 8, with no carry-over effect, multivariate normal distribution data according to effect difference was produced. Paired t-test, mixed effect model and mixed effect model based on difference were used. Type [ error,power,mean error（ME） ,ab- solute mean Error（ AE）, Root mean square error（RMSE） were used to e- valuate the models. Results The mean of estimated value in all modelswere very proximity effect difference. ME, AE,RMSE of estimated value were small. Type I error of all models were approximately equal to 0. 05, except model 7. With the increase of difference effect, power of all models increased. When difference effect between two groups was small （ 〈 1.0 ）, model 5 had the maximum power,model 2,model 3 and model 4 had the minimum power. When difference effect between two groups was large（ ≥1.0 ）, power of all models was almost the same （ 〈 0. 010 ）. Conclusion Mixed effect model is more suitable than paired t-test to analyze n-of-1 data of correlation. The effect of mixed effects model is better than mixed effects model based on difference. The optimal model was mixed effects model based on CS structure.