Monte Carlo-based data snooping with application to a geodetic network

Data snooping is one of the best established methods of gross error detection in geodetic data analysis. Since it is based on hypothesis testing, it requires the choice of levels of error probability. This choice is often, to some degree, arbitrary. If the levels chosen are too high, we run the risk of losing many good measurements that are not actually contaminated by gross errors. If the levels chosen are too low, we run the risk of leaving gross errors undetected. We propose to choose levels of error probability such that the desired parameters are best estimated in some sense. This can be done using the Monte Carlo method. We applied this procedure to a geodetic precision network from construction of a diversion tunnel. Depending on the stochastic model of the measurement process, we observed a gain of such an optimal choice of a few percent of the mean point standard deviation. This comes at a price of considerable computer time consumption. Even on a fast computer, a typical computation of a medium-sized geodetic network may take several minutes.