Testing a palaeomagnetic study for the averaging of secular variation

Summary. It is shown that the interval estimate for K given by Cox (1969) is inappropriate as a test for determining whether secular variation has been averaged out in a particular study. The appropriate test is presented in two formulations, one for use with χ² tables and the other for use with F tables.
The well known secular variation models are investigated and it is shown that Model A due to Irving & Ward (1964) is, statistically, the most efficient predictor. Finally, for the 83 palaeomagnetic results selected by Brock (1971), it is shown that the most efficient predictor for the secular variation K is k '= 18.1 (1 + 3 sin²λ), where λ is the palaeolatitude, and the uncertainty in this predicted value is discussed.