Biology Reference
In-Depth Information
placed at the base of the ischial spine and the tip of the piriform tubercle (
Figure 14.2
). The
semilandmarks were placed along this arch using a “fan” guideline produced using the
MakeFan program. Similarly, two landmarks and 35 semilandmarks were used to capture
information about the ischiopubic region. One was placed at the intersection of the upper
edge of the pubis with the perpendicular line that reaches the uppermost point of the
obdurator groove. The second was placed at the intersection between the external margin of
the ischium and the inferior border of the acetabulum. Thirty-five semilandmarks were then
placed along the curve between these points. The placement of the first landmark on the
ischiopubic region was difficult in this study as it required an estimated location based on a
perpendicular line due to a lack of a clearly defined landmark in this region.
After landmark and semilandmark placement was completed, a semilandmark align-
ment procedure was done, using distance minimization on a mean form estimated using
Generalized Procrustes Analysis methods (which also iteratively estimates the mean). PCA
analysis of each data set showed substantial evidence of segregration of males and females
along the first axis (58.2% of variance in the sciatic notch data, and 56.9% of the variance
in the ischiopubic complex). These results were compared to those obtained from discrimi-
nant analysis and
k
-means clustering (
MacQueen, 1967
). Each of these analyses was
performed on the same shape data as the PCA and on a combined size
shape data set
(Procrustes form space). For the discriminant analysis (similar to CVA, but with only a
single axis), correct classification rates were obtained using a resampling-based cross-
validation analysis. In the
k
-means clustering method, the specimens are divided into
1
FIGURE 14.2
Landmark and semilandmark placement
on the sciatic notch (a) and the ischiopubic complex (b)
(From
Gonzalez et al., 2009
).
