How to estimate the minimum power of the test and bound values for the confidence interval of Data Snooping procedure

Data Snooping (DS) is the best-established method to identify gross errors (outliers) in geodetic data analysis, with a given probability. The power of the test is the probability of DS correctly identifying a gross error, while the confidence interval is the probability of DS not rejecting an observation uncontaminated by gross error. In practice, the power of the test is always unknown. Thus, the objective of this paper is to present a theoretical review of how to determine the minimum power of the test, and bound values for the confidence interval of the DS procedure in an n-dimensional scenario, i.e., considering all observations involved. Along with the theoretical review, a numerical example involving a simulated leveling network is presented. The results obtained in the experiments agreed with the previously calculated theoretical values, i.e., the revised methodology showed satisfactory performance in practice. The example also shows the importance of the revised methodology in the planning stage (or pre-analysis) of geodetic networks