Techniques for dealing with discordant observations

Abstract

To test any theory such as theories of motion–Newtonian or relativistic–of solar system objects, one must compare the predictions of theory with observation. But discordant observations habitually plague the reducer of astronomical data. To alleviate the baleful effects, particularly harmful when the observations are reduced by the method of least squares, of discordant data investigators almost invariably reject observations whose corresponding (0-C)'s or post-solution residuals exceed a cutoff. But techniques that are insensitive to the assumption that the observational errors are normally distributed, called robust estimation in the literature, have also been developed.