Abstract:
In this note we check the interesting recent results of Gomez Portugal Aguilar {it et al} cite{aguilar00}. These authors fit a simple random walk with Gaussian increments to the radiocarbon calibration curve. They suggest that the posterior standard deviation of the calibration curve is smaller between years where observations were made than it is at those years. We find that in contrast to their result the posterior standard deviations of the calibration curve bow out slightly between observation points in our estimates, for a model of the kind they describe. In earlier work Christen~cite{christen94} conditions the random walk to visit the measured calibration values. Following cite{aguilar00}, we consider the unconditioned posterior distribution for calibration curves, and thereby improve slightly on the reconstructions of cite{christen94}.
