Bayesian approach to smoothing palaeomagnetic data using ABIC

Bayesian spline regression for 1-D time series is useful for smoothing palaeointensity data. This method, however, is not applicable to directional data composed of inclination and declination. In the present study a new spline regression, based on Bayesian statistics, is developed for smoothing a time series of unit vectors subjected to Fisher distribution. An optimal smoothing factor is determined by minimizing a Bayesian information criterion (ABIC) in this new method. These methods are applied to archaeomagnetic data in Japan, and the X, Y and Z components are estimated in the absolute scale for the time span of 200–2000 yr BP.