Biology Reference
In-Depth Information
available (see the wiki cited above), and all of the code used for this chapter is available
online
2
.
Second, we make extensive use of simulated datasets rather than actual osteological data
in this chapter. The advantage of using simulated data is that it is essentially known-age and
known-sex because we created it, and we consequently can check that the demographic
analyses are producing the correct answers. While we might accomplish the same thing
with osteological data from known-age, known-sex collections, such collections are rare
and often do not represent natural population profiles. Within our datasets, the osteological
variables used to estimate sex or age are simulated to follow what is known about the statis-
tical dependency of these traits on the demographic variables (actual sex and age) as realis-
tically as possible. The simulations of the sex or age-at-death distributions that generate the
dependent osteological traits also are designed to demonstrate particular properties of the
analytical methods. All of the simulated datasets, as well as the code that generated them
2
.
We believe firmly in learning by doing, and hope you will take advantage of the online mate-
rials to work through and play with the examples.
STARTING DATA FOR DEMOGRAPHIC ANALYSIS OF SKELETAL
SAMPLES
If individual age or sex estimates cannot be produced until after the demographic analyses
have been performed (for reasons that will become clear in subsequent sections), then what
information from the skeletal sample initiates the demographic analyses?
Hoppa and Vaupel
(2002b)
state that when working with age and ordered categorical data (such as pubic
symphyseal scores), “the information that osteologists have regarding age and stages
pertains to the probability of being in a specific stage given age,
PrðcjaÞ
.” Generalizing this
statement, an osteologist needs two things in order to conduct a demographic analysis: (1)
recorded information on individual sex or age indicators from each skeleton, and (2)
a way to evaluate how these indicators depend, or are conditional, on known sex and/or
age. The skeletal sample with unknown ages and/or sexes for which we wish to estimate
population-level demographic parameters and individual ages and/or sexes forms a target
sample (
Konigsberg and Frankenberg, 1992
), and it is from this sample that we will need to
have the recorded information on individual sex and/or age indicators. It is imperative in
constructing the demographic analysis that we have this basic observational information
from each skeleton, so for example we will need the “scores” for each indicator. A proper
demographic analysis of the target sample cannot begin with age estimates for each skeleton;
it must begin with the basic observations that (traditionally) an osteologist would consider in
making an age estimate. In order to effectively use observations on indicators from the target
sample we must understand how these indicators are dependent on individual demographic
variables (age and/or sex).
Generalizing Hoppa and Vaupel's
PrðcjaÞ
probability, we want to find the probability of
various indicator states or values dependent on known individual age and/or sex. This
information must come from a known age and/or sex reference sample (
Konigsberg and
