How many paleointensity determinations are required from a single lava flow to constitute a reliable average?

Abstract

Statistical analysis has been undertaken to investigate the minimum number of paleointensity estimates required to allow certainty (at the 95% confidence limit) that the mean is within 10% of the implied true mean for an entire lava flow. The data used for the analysis was provided by previous, paleointensity studies which used three different methods (Thellier-Coe, Shaw, and microwave Thellier) on 19 or more samples from three different basalt lava flows. The results strongly suggest that future paleointensity studies need to produce many more estimates per lava flow than is currently the norm in order to ensure that the calculated mean paleointensity does not differ significantly from the true flow mean. Furthermore, they indicate that a low measured dispersion for a small number of estimates is not a safeguard against misrepresentation of the flow mean paleointensity. Finally, we provide quantitative evidence that paleointensity studies should always attempt to maximise vertical coverage of lava flows to avoid biasing of the mean.