Geomagnetic lunar analysis—the estimation of errors

Summary. It has recently been proposed that a reliable estimate of the errors of tidal harmonics present in geomagnetic data can be obtained by the following method: (1) randomly assign each observation to one of 10 subsets, (2) determine the tidal harmonics separately for each subset, and (3) calculate the standard deviation of the mean from the scatter among the 10 determinations. This method is valid if the subsets are statistically independent, but will lead to an underestimate of the errors if they are not. Here we show that, for real geomagnetic data, the assumption of statistical independence is valid.