基于SAS对缺失数据的处理——以新药试验为例

Treatment of Missing Values in Drug Experiments by Using SAS

通过对新药试验的数据集中缺失数据进行分析,总结缺失数据对新药研究分析结果的影响,进而比较不同补缺方法的优缺点.利用SAS9.4软件对新药试验的数据集进行统计分析,分别采用LOCF补缺法、均值补缺法、分组均值补缺法、MCMC补缺法、FCS补缺法、monotone补缺法六种方法对缺失数据集进行统计分析,并比较数据填补前后的差异及分析结果的稳定性.由于新药试验数据集的病例脱落比例不是很高,故几种处理方法的分析结论都比较接近,但MCMC填补法的效果相对其他几种的填补方法更为稳定.

Influence of missing values on new drug trial is analyzed by comparison with different imputation methods. Statistical data on new drugs are treated by using six imputation methods in SASg. 4 such as the LOCF method, mean-value method, group-mean method, MCMC method, fcs method and monotone method. Stability and differences of statistical data before imputation are compared with them after imputation. Results by using above mentioned methods appear similarity due to a few of expulsion in data of new drug trials. It is noted that the MCMC method performs better stability than others do.