They don't look different, but they're not the same: formal resemblance and interpretive disparity in the construction of temporal frequency distributions

TL;DR

This paper examines various protocols for constructing temporal frequency distributions in archaeological and paleontological data, highlighting their formal similarities but fundamental interpretive differences, and advocates for uncertainty propagation techniques.

Contribution

It introduces the importance of propagating multiple sources of uncertainty in temporal frequency distribution construction, emphasizing the interpretive disparities among similar methods.

Findings

Protocols are algebraically similar but interpretively different.

Different sources of uncertainty require distinct handling.

Propagation-of-uncertainty techniques improve tfd reliability.

Abstract

In archaeological and paleontological demographic temporal frequency analysis (dTFA), a handful of protocols for generating temporal frequency distributions (tfds) have emerged based on the aggregation not of single-point timestamps but instead of constituent temporal distributions, including probability summation, kernel density estimation, and human occupation index calculation. While these protocols bear a striking algebraic resemblance to one another, they are motivated by the desire to contain fundamentally different sources of uncertainty, leading to detailed differences in procedure as well as fundamental differences in the interpretation of the resulting tfd. Rather than assuming that one technique can fulfil dual purposes based on its formal resemblance with another, the joint containment of multiple sources of uncertainty therefore warrants the adoption of propagation-of-uncertainty techniques in tfd construction.