Figure 5. Effect of the number of samples per paleomagnetic site (n) on the a) angular distance of the acquired paleopole to the reference pole with n=7 (or 6, in case of Antarctica), and the b) A95,c) K, and d) S of the acquired paleopole. Every datapoint represents 1000 simulations, with error bars indicating the 95% confidence interval.
4.3 Effect of discarding outliers on site level
One potential benefit of having multiple samples per site is to discard outliers, commonly done in the paleomagnetic community based on the ‘expert eye’ of the interpreter. Here, we objectively studied the effect of discarding outliers on site level, by eliminating the farthest outlying sample(s) from the site. We did this three times, each time discarding the sample furthest from the recalculated site mean. Interestingly, we find generally no effect of eliminating within-site outliers on mean paleopole position and its A95, K, and S (Figure 6), and for the Norway dataset, the between-site scatter even increases. This is because VGP positions do not systematically shift towards or away from the paleopole (Figures S2-S5). As a result, decreasing the within-site scatter by removing outlying directions has little effect on the between-site dispersion of VGPs.