Biology Reference
In-Depth Information
and Hens, 1998; Ubelaker and Volk, 2002; Walker, 2008 ), which has great potential for future
research. Walker (2008) tested multiple types of discriminant function analysis using ranked
ordinal scores of visually assessed features of the skull: mental eminence, orbit margin,
glabellar region, nuchal area, and mastoid process. He found that logistic regression discrim-
inant function analysis was the best statistical procedure to minimize misclassifications and
sex biases of classification; he achieved 88% accuracy. This method is extremely population
specific due to the aforementioned population variation in robusticity and gracility ( Walker,
2008 ). Discriminant function analysis will be described later in this chapter under the section
on metric sex estimation.
When beginning to test a new (or existing) trait for sex assessment, a useful process is to
order all of the variation as a continuum, a process known as seriation. To seriate skulls
(or any bones for that matter), you will select a specific trait of interest and then organize
the skulls (that are clearly labelled with case numbers to preserve provenience) from those
with absence or the least expression of the trait to those with greatest expression of the trait.
“When sexing only skulls, always use the entire population under study” ( White et al., 2012 ).
White and colleagues state that the most accurate method of sex estimation is via the seriation
method within a single population. Through training and experience, the seriation method
yields correct sorting 80
90% of the time ( White et al., 2012 ).
There is always overlap between males and females near the center of the distribution
(often called the zone of overlap between the sexes), which makes sex estimation indeter-
minate within this middle range. Accuracy for those that are “certainly” one sex or the
other (i.e., those outside the zone of overlap) is 100% in tests ( Asala, 2002 ). See Figure
11.1 from Konigsberg and Frankenberg (Chapter 11) this volume, for a graphic depiction
of this zone of overlap in sexual characters. If not using a sophisticated statistical analysis,
this zone of overlap can unfortunately often include the majority of the population, which
is why statistical analysis is preferred even for sex assessment of nonmetric traits. If the
zone of overlap only includes a few individuals, the likelihood of the sex assessment can
still be very high. One drawback is that this seriation method relies on a larger sample
size. A situation in which there is more likely a sample large enough for seriation is found
in archaeological samples, less so in forensic cases ( White et al., 2012 ). However, large
samples of individuals from a single population have also been recovered in forensic
human rights investigations of mass graves, to which this technique could be applied for
sex assessment.
e
METRIC SEX ESTIMATION
Metric sex estimation from the skeleton involves the quantitative analysis of measurable
sexually dimorphic traits, such as the femoral head diameter or scapular height. Sex estima-
tion using metric analysis has only recently become the standard in both forensic anthro-
pology and bioarchaeology. Almost a century ago, Pearson (1915) first suggested that
postcranial metrics be used for sex estimation. It has taken a long time for this to catch on,
due to the speed and ease of visual sex assessment. Metric analysis, however, typically
involves less subjectivity and lower inter- and intraobserver error ( Adams and Byrd, 2002;
Spradley and Jantz, 2011 ).
Search WWH ::




Custom Search