Morphometric Data - An Invariant Approach to Statistical Analysis of Shapes - page 40

Table 2.2. Landmarks used in Corner et al. (1992) with their abbre-
viations
1) Nasion (NAS)
2) Nasale (NAL)
3) Right Zygomaxiilare inferior (ZMI)
4) Right Zygomaxiilare superior (ZMS)
5) Intradentale superior (IDS)
6) Posterior Nasal Spine:Vomer-Palatine intersec-
tion (PNS)
7) Right Maxilliary Tuberosity:Maxilliary-
Palatine junction (MXT)
8) Vomer-Sphenoid junction (VSJ)
9) Right Pterion Anterior: Zygo-Spheno-Frontal
junction (PTA)
10) Bregma (BRG)
11) Right Carotid Canal: center of canal (CAR)
Chapter 3 for such an approach), the impact of measurement error on
the reliability of the distance data varies with the distance. The pro-
portion of error is greater over small distances than it is over large
distances. For example, in the measurement error study described
above, the linear distances between any two landmarks ranged from
0.72 to 7.6 cms. Even if the maximum error along any of the axes
(0.451 mm) occurred at a landmark defining an endpoint of the small-
est linear distance, the error constitutes only 16% of that distance. For
larger distances, the error is negligible. Consequently, the imprecision
that can be tolerated in a study varies with the measures being used
and comparisons being made (Kohn and Cheverud, 1992). Third, mea-
surement error can systematically vary among the landmarks studied.
For example, in this particular study larger measurement error was
found at landmarks that mark the extreme ends of sutures located on
an osseous edge (e.g., posterior nasal spine located at the distal end of
the internasal suture; zygomaxillare superior located at the zygomax-
illary suture along the orbital rim). Finally, if measurement error
seems to be excessive, the following simple steps can be taken to
reduce its impact. Data should always be collected at least twice. If nei-
ther trial contains large measurement error, the trials can be averaged
to represent the specimen in analysis. When particular landmarks
show increased error, and the landmark is needed in analysis, we sug-
gest that the problematic landmark coordinates be collected