Everything is not yet lost: Modeling taphonomic bias in a Bayesian survival-analytic framework

Summary

The time-transgressive loss of archaeological, paleontological, and other geological deposits to destructive geomorphic processes has been parametrically modeled by T. Surovell and colleagues (2009), with minor revisions offered by A. Williams (2012). We expand on these modeling efforts in an explicitly survival-analytic framework, employing analytical techniques tailored to the study of time-to-event processes and data. First we show that Surovell and colleagues’ model is in fact a reduced (two-parameter) form of the Lomax survival model. Then we refit this reduced Lomax model through Bayesian inference to the same terrestrial record of global volcanic deposits employed by Surovell and colleagues and by Williams, basing our parameter estimates on the maximum a posteriori (MAP) estimator and highest-probability credible region. We further expand these modeling efforts by fitting both the two-parameter Weibull model and its reduced (one-parameter) negative exponential form to the same volcanic data set. Model selection based on the Bayesian information criterion (BIC) favors the Weibull model over both the negative exponential and the reduced Lomax model. Implications of the fitted Weibull model for the robust correction of archaeological temporal frequency distributions (tfds) are considered.

This Resource is Part of the Following Collections

• Society for American Archaeology 81st Annual Meeting, Orlando, FL (2016)

Keywords