Improved radiocarbon age estimation using the bootstrap.

Abstract

This paper proposes the use of the statistical bootstrap technique as an aid in combining radiocarbon date estimates. The rationale for the use of the bootstrap is the theoretical result that, even if individual date estimates are normally distributed, their combination by the usual formula results in a random quantity that is not normal but rather a mixture of distributions. The bootstrap is a non-parametric, computer-intensive technique. This technique can better estimate the actual distribution of the combined age, leading to more precise confidence intervals. While the bootstrap cannot solve the multiple-intercepts problem in calibration, it can nonetheless lead to better estimates. The benefits of using the bootstrap are especially noticeable when sample sizes are small (as is the case in other applications of this technique).