Morphometric Data - An Invariant Approach to Statistical Analysis of Shapes - page 39

error associated with each of the landmarks along the three major
axes. These are reported as standard deviations and not as variances
(see Table 2.2 a nd Figure 1.2 for landmark identification). Each observ-
er shows a high consistency in locating the landmarks. The standard
deviation along any of the three axes ranges from 0.146 mm to 0.451
mm. The consistently low variation in coordinate values across the
three dimensions indicates that no single direction is particularly
prone to error for any of the landmarks considered here. Table 2.1 also
indicates that Observer 2 (who was more experienced with the use of
the machine and the anatomy of the skull) achieved a slightly better
measurement error than Observer 1.

Table 2.1. Measurement error for 3Space digitizer (Corner et al.,
1992). These are standard deviations (in mm ) for the twenty repeated
measurements of a fixed Macaca fascicularis skull. Larger values
indicate larger measurement error or lack of precision. Definitions for
the landmark abbreviations are given in Table 2.2 and some of the
landmarks are pictured in Figure 1.2 .

Landmark Observer 1 Observer 2

X Y Z X Y Z

NAS .249 .379 .248 .173 .256 .195

NAL .248 .451 .378 .361 .229 .304

ZMI .163 .208 .338 .154 .143 .336

ZMS .246 .385 .417 .199 .280 .446

IDS .297 .395 .259 .287 .170 .205

PNS .201 .281 .270 .228 .350 .373

MXT .211 .180 .289 .300 .157 .446

VSJ .226 .146 .253 .294 .166 .383

PTA .224 .258 .343 .196 .216 .212

BRG .165 .197 .204 .194 .219 .330

CAR .200 .234 .239 .335 .201 .278

This study enables us to provide the following recommendations for
the use of 3D digitizers as data collection devices. First, the researcher
should become very familiar with both the specimens under study and
the data collection device. We have found data collection practice ses-
sions to be very useful. Second, when landmark data are collected but
linear distances are computed for further statistical analysis (See