Biology Reference
In-Depth Information
0.05, carefully consider the measurements entered and the populations (reference samples)
included in the analysis.
Case Study: Using FORDISC
Identifying the appropriate reference sample is one of the more daunting aspects of
ancestry assessment using FD3. FD3 has two major samples to which an unknown may
be compared. The first is the Forensic Databank (FDB) (
Jantz and Moore-Jansen, 1988
),
which had approximately 3400 cases and growing as of early 2011 (
Ousley et al., 2011
).
The FDB consists of identified individuals from forensic cases originating predominantly
from the United States. The second is the Howells database of 2504 individuals from 28 pop-
ulations around the world compiled by W.W. Howells. When doing an analysis using
FORDISC, the user chooses to which groups the unknown cranium should be compared.
However, the challenge is in deciding at what point one group or another should be
excluded from the comparison. The general consensus is to begin with a broad approach.
In FD3, this would include all possible groups in the first analysis, and then, based on those
results, removal of populations that are not probable. For example, if your results in the first
analysis suggest a male individual, with all values being highest for males regardless of the
population, then all females should be removed and the analysis processed again.
For the sake of example, let us assume that in the second analysis the values in
Table 5.1
are
obtained. Clearly, this individual is not a white male (D
2
0.000). In fact, it
is highly unlikely (improbable) that the cranium in the example above belongs to Hispanic
males, Amerindians, or Black males, based on the low posterior and typicality probabilities.
Once those groups are removed and the analysis is recalculated, we get the following results
in
Table 5.2
.
Although the classification (Vietnamese male) has not changed, we have narrowed down
the potential list of ancestral groups to four populations. Again, we could continue to narrow
the populations down to just two groups (as suggested in the FD3 manual). However, for our
purposes we can feel confident that this individual is most likely Vietnamese, though not to
the exclusion of other reasonable possibilities. Wait a minute! That doesn't seem good
enough. If you are like us, you will see that this is not enough information to make a final
estimation of ancestry. Even if we reduce the number of variables (
Table 5.3
using a stepwise
selection
e
see the FD3 manual for a detailed discussion of this process) we are no closer to
a final determination.
In fact, these results further muddy the issue because now the VM and GTM results are
nearly identical. This is not an uncommon situation and it clearly demonstrates that a proper
understanding of human variation, metric analysis, and nonmetric traits is necessary not
only to correctly assess ancestry, but also to correctly interpret FD3 results and properly select
reference samples.
So what are we to make of the example case described above? All of the assumptions for
discriminant function are met, so the DFA appears to be performing well. Other chapters in
this volume describe the importance of context when interpreting results from skeletal
analyses. Perhaps the context (i.e., situation in which they were found) of these remains
can assist in making our final decision.
This example was taken from the FORDISC 3.0 help file (
Jantz and Ousley, 2005
) and is
identified therein as Example 2. The measurements are from a University of Tennessee
¼
28.6; Post. Prob.
¼
