In Conventional Deformation Analysis (CDA), at least two different epochs are adjusted by using the Least Squares Estimation (LSE) method and compared statistically. The effect of the geometry of the network is an essential part of the adjustment model and the LSE method smears the effects of the displaced point over the other nondisplaced points. In this study, to remove these spoiling effects and to increase the reliability of the deformation analysis, a new approach is introduced. This approach depends on the analysis of the differences between observations of the two epochs, and also considers the principles of the model error approach. All possible combinations of the differences of the observations are considered as model errors in Gauss-Markov model and the estimated model error for the combination, that has the smallest variance, is compared with a critical value to answer the question whether it is significant or not. To compare the results of the new approach with the CDA, the Monte Carlo simulation technique and mean success rate are used in a leveling network. As a consequence, according to the simulation results, the new approach is better than the CDA by 7.6% and 9.7% for one and two displaced points, respectively, when the deformation network is designed as a subnetwork.

在传统变形分析（CDA）中，使用最小二乘估计（LSE）方法调整至少两个不同的历元并进行统计比较。网络几何形状的影响是平差模型的重要组成部分，LSE 方法将位移点的影响涂抹到其他非位移点上。在本研究中，为了消除这些破坏效应并提高变形分析的可靠性，引入了一种新方法。这种方法依赖于对两个历元观测值之间差异的分析，并且还考虑了模型误差方法的原理。观测值差异的所有可能组合都被视为高斯-马尔可夫模型中的模型误差，并且将方差最小的组合的估计模型误差与临界值进行比较，以回答其是否显着的问题。为了将新方法与 CDA 的结果进行比较，在水准网络中使用了蒙特卡罗模拟技术和平均成功率。因此，根据仿真结果，当变形网络设计为子网络时，对于一个和两个位移点，新方法分别比 CDA 好 7.6% 和 9.7%。