ACCOUNTING FOR DATING UNCERTAINTY IN PALEONTOLOGICAL ANALYSES

In paleontological analyses, it is common for ages of fossil occurrences to be known only as geochronological units (e.g., Famennian stage) rather than as exact numerical values (e.g., 364.9 Ma). This uncertainty can make it difficult to apply quantitative and statistical analyses that require exact numerical values, such as the estimation of times of origination and extinction, reconstruction of phylogenetic trees, or the calculation of evolutionary rates. A common solution is to randomly choose a numerical age from a uniform distribution bounded by the beginning and end of the corresponding time interval. After a randomized age has been drawn for each fossil occurrence, the analysis can be applied to this dataset of numerical values. The randomization and analysis are then repeated a large number of times to account for variability in the randomized ages. If each analysis yields a confidence interval for some parameter of interest, this procedure will result in multiple confidence intervals. However, it is not obvious how to combine these multiple confidence intervals into a single answer. We investigate how to deal with multiple confidence intervals arising from such randomization procedures, and also explore the effect of dating uncertainty on paleontological inferences.