Biology Reference
In-Depth Information
Frankenberg, 1992
), or its equivalent. The reference sample should be a sample with known
demographic information that is appropriate for application to the target sample. Unfortu-
nately, the acceptance of a reference sample as being “appropriate” is often more of a leap of
faith than anything else. For example, Shackelford et al. (2012) have applied data on dental
formation and eruption from known age twentieth century children to estimate ages-at-
deathforNeandertalandearlymodernhumanfossils.Inanactualanalysisitmaynotbe
necessary to have access to a reference sample, provided that summary statistics of the rela-
tionship between osteological traits and known age and/or sex have been published in
a useable format (for example, the mean and standard deviation of each indicator for each
known age and/or sex). It also may be possible to combine information from different sour-
cessuchascombiningsummarystatisticsonfetal long bone growth from ultrasound exam-
inations with summary statistics on postnatal long bone growth from radiographs.
What is absolutely critical is that we have information on the indicator states given
known age and/or sex. Much of the literature (see for example
Todd (1920)
,
Stewart
(1948), McKern and Stewart (1957), Thompson (1979), Meindl and Lovejoy (1985)
),
because it focuses on estimating age and/or sex for individual skeletons, does not provide
such information in a useable format. This early literature has instead provided summa-
ries of the distribution of age (such as mean ages, standard deviations of age, or age
ranges) within stages. Similarly, when trying to estimate sex we need information on
the distribution of an osteological “indicator” of sex within known sex, and not the distri-
bution of sex against the indicator.
As a brief example, presume that we have assigned sexes based on mastoid process size in
a sample of known sex individuals, and that we have done this assignment without reference
to the known sexes. To simplify the example, we will say that we treat the mastoid process
size as a binary variable, so that we only assign sexes of “F” or “M.” Now we go back and
look at the two-by-two table of known sex (male or female) against the sex assigned from
the mastoid process. The useful information in this table is
Pð
}
M
}
jmaleÞ
and
Pð
}
F
}
jfemaleÞ
,
where
Pð
}
M
}
jmaleÞ
is read as “the probability of scoring a mastoid process as being male
given that the individual is an actual male.” Complementary values can be obtained by
subtraction (for example,
Pð
}
F
}
jmaleÞ¼1 Pð
}
M
}
jmaleÞ
). The information contained in
the “transposed conditionals”
Pðmalej
}
M
}
Þ
and
Pðfemalej
}
F
}
Þ
is much less useful. This subtle
distinction will be easier to demonstrate in concrete examples such as we present below.
ESTIMATION OF SEX AND OF THE SEX RATIO
While it may seem odd to refer to “estimation” of sex, it should be clear from Moore
(Chapter 4), this volume, that the sex of skeletons can be treated as known only under certain
circumstances. Specifically, if DNA sexing (
Hummel and Herrmann, 1991; Stone et al., 1996;
Faerman et al., 1998; Mays and Faerman, 2001; Matheson and Loy, 2001; Schmidt et al., 2003;
Arnay-de-la-Rosa et al., 2007
;
De La Cruz, 2008
;
Gibbon et al., 2009
) has been applied, then
the sex of individual skeletons can be treated as known. Additionally, if the
Phenice (1969)
pubic bone characteristics are observable and unambiguous, then sex can be treated as nearly
known, with correct identification of sex ranging between 95% and 98.5% (
Kelley, 1978;
Sutherland and Suchey, 1991; Konigsberg et al., 2002
). One study (
Lovell, 1989
) did give
