Biology Reference
In-Depth Information
Unlike the
Purkait (2005)
method discussed above, the Albanese et al. method (2008) was
developed to measure the size and angle of the femoral neck. This method was developed on
more than 300 individuals from the Terry Collection and analyzed using the statistical method
of logistic regression. The extremely high accuracy achieved using this method was between
95% and 97%.
Albanese et al. (2008)
claimed that their method is not population specific, but
rather relies on biomechanical differences in males and females. The authors explain that they
used logistic regression instead of the discriminant function approach for a number of
reasons, with one being that logistic regression performs as well as discriminant function anal-
ysis, but has fewer assumptions (e.g., does not assume a normal distribution) (
Albanese et al.,
2008
). A suggested research project could test the accuracy of the Albanese method on other
population samples to determine the verity of the universal application of the approach.
One earlier study by
I¸can and Miller-Shaivitz (1986)
used discriminant function analysis
to test whether the femur and tibia are equally dimorphic. They found that the femur and
tibia are in fact equally effective for sex estimation in European Americans from the
Hamann-Todd Human Osteological Collection,
10
but that only the femur was useful for
estimation of sex in African Americans from that sample.
I¸can and Miller-Shaivitz (1986)
concluded that the epiphyses were better than the shaft for dimorphism due to the muscle
insertion sites, with accuracy rates between 83% and 94%.
Using a traditional metric method, Asala tested the efficiency of the demarking point of
the femoral head to estimate sex (
Asala, 2001, 2002
). The vertical and transverse diameters
of the femoral head were measured and averaged for each sex. The average was then taken
of the male and female averages to arrive at a demarking point (
Bidmos and Dayal, 2004
). The
results revealed that the male femoral head diameters were significantly greater than those of
the females. Using this simple method of sorting without discriminant function analysis can
have very low overall success (only 32%) due to the zone of overlap between the sexes, but
the 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
). A sectioning point is based on the assumption that the
standard deviations for both sexes is equal and that the sex ratio is exactly one to one, which
is not necessarily the case. You need to first have an idea of the sex ratio. Without a sex ratio,
you can estimate a ratio using the statistical procedure called maximum likelihood estima-
tion (MLE). You can then use a Bayesian analysis to find the probability that an individual
with a given measurement is male or female. For a more in depth explanation to this
approach, refer to the chapter by
Konigsberg and Frankenberg
on demography in this
volume (Chapter 11).
Asala and colleagues (2004) also conducted discriminant function analysis on 220 femora
from South African Blacks using just the proximal and distal segments to calculate the sex
estimation accuracy for fragmentary femora. They found multivariate analyses were better
than univariate, with the proximal end of the femur (85.1%) more effective in sexing than
the distal end (82.7%). These results were interpreted as potentially due to the role of the
proximal femur in transmitting body weight. A more complex statistical analysis from
three-dimensional computer models created from computed tomographic (CT) scans reveals
an internal rotation of the distal femur in females, perhaps as a biomechanical means of
10
This collection is curated by the Cleveland Museum of Natural History in Cleveland, Ohio.
