Albyn Jones
I plan to prepare a case study on carbon dating.
Raw carbon dates (actually ratios of C14 to C12 concentration) are
measured with error. In addition, the atmospheric ratio is variable
over time, so one can't simply extrapolate back using the half-life of
C14: there is a calibration problem. A calibration curve based on
dendrochronological studies (ie trees of known ages) allows the
construction of the calibration curve. Finally, the computation of
a calendar year for the sample can be done by maximum likelihood
of Bayesian methods. The problem can be complicated by the existence
of multiple dates from the same tree with intervals known by counting
tree rings, or by multiple samples with known order based on stratigraphic
evidence.