Some adjustments are made to design weights to reduce the negative effects of non-response and out-of-scope problems. The calibration approach is a weighting process that agrees with the known population values by using auxiliary information. In this study, alternative calibration approaches and weight trimming process that can be used in large data sets with extreme weights and different correlation structures were analysed. In addition, the effect of the correlation structure of auxiliary variables on the efficiency of the calibration estimators was investigated by a simulation study. The 2017 Annual Industry and Service Statistics data were used in the simulation study and it was seen that restricted calibration estimators were more efficient than the generalized regression estimator in estimating the variables with a high variance such as turnover. Especially in small sample fractions, we recommend the application of restricted calibration estimators, as they are more efficient than the weight trimming in solving the negative and less than one weights problem encountered after the calibration process.

对设计权重进行了一些调整，以减少不响应和范围外问题的负面影响。校准方法是通过使用辅助信息与已知总体值一致的加权过程。在这项研究中，分析了可用于具有极端权重和不同相关结构的大型数据集的替代校准方法和权重修整过程。此外，通过模拟研究研究了辅助变量的相关结构对校准估计器效率的影响。在模拟研究中使用了 2017 年度工业和服务统计数据，可以看出，在估计营业额等高方差变量时，受限校准估计量比广义回归估计量更有效。特别是在小样本部分，我们建议应用受限校准估计器，因为它们在解决校准过程后遇到的负数和少于一个权重问题方面比权重微调更有效。