Biology Reference
In-Depth Information
American sample the mean age-at-death estimated from pubic symphyses differed by
10 years from the mean age-at-death estimate from cranial sutures. Her findings
resulted in the caution that the skeleton was an entire unit during life so “no one
age indicator is adequate” (
Brooks, 1955
:588).
Lovejoy et al. (1985b)
presented the Multifactorial Summary Age method, which utilizes
the pubic symphyseal face, auricular surface, radiographs of the proximal femur, dental
wear, and suture closure. Their evaluation of skeletal aging methods summarizes that
most methods are fairly inaccurate and can be gravely affected by interobserver error. This
technique is intended for researchers engaged in paleodemography, so its applicability to
individual specimens in forensic situations is questionable; however, it could be a plausible
method in analysis of mass disasters or mass graves, although to this author's knowledge it is
not used for such situations.
After the specimens are seriated according to
Lovejoy et al. (1985b)
, each aging indicator is
applied separately to each specimen and the scores are used to make an intercorrelation
matrix,
13
which is then subjected to principal components analysis (a multivariate statistical
procedure that combines several variables into a single variable for analysis). The correlation
of each indicator with the first principal component is then taken as that variable's weight.
Any factor with a correlation of at least r
0.70 is incorporated into the model. The weighted
average of each indicator then becomes the final age estimate (the “summary age”) for each
individual.
Over the past few decades a few more authors have attempted multifactorial methods
(e.g.,
Saunders, 1992; Baccino and Zerilli, 1997; Baccino et al., 1999
). It is clear from the
literature that the use of multiple age indicators for the determination of age-at-death is
ideal, but a scientific, quantitative, and easily applied method for combining the data is
lacking. A comprehensive approach to aging is needed to simplify, and more importantly,
standardize age-at-death estimation from skeletal remains. Further research is needed in
this area.
Recently, paleodemographers have been at the forefront of multifactorial age-at-death
estimation. Boldsen and colleagues (2002) developed a computer program (ADBOU)
that collects data on multiple skeletal indicators scored as discrete ordinal phases
and uses Bayesian inference to calculate the posterior probability density and estimate
age-at-death. Unfortunately, tests of the ADBOU program have found it only moder-
ately effective (
Bethard, 2005; Uhl, 2008; Milner and Boldsen, 2012
), in part because
the trait scoring departs from the methods (e.g., Suchey
e
Brooks) to which many oste-
ologists are accustomed. Without extensive practice, intra- and interobserver error can
be problematic. Further, the ADBOU program comes with only a small choice of prior
age-at-death distributions "hardwired" into the program. Bayesian analyses rely on
these prior probabilities, together with the osteological data, to estimate ages at death
for individual cases.
¼
13
The authors use the intercorrelation matrix as a means to generate the principal components, the first
component of which is assumed to represent chronological age. The correlation of each indicator to that
principal component is then used to weigh its contribution to a final multifactorial age estimate for an
individual.
