Testing repeatability and error of coordinate landmark data acquired from crania.

This study evaluates issues of precision, repeatability, and validation in three-dimensional (3D) landmark coordinates. Two observers collected 19 homologous cranial landmarks from three skulls during three separate digitizing sessions. Statistical analysis was conducted on the 171 interlandmark distances (ILDs) derived from the original coordinate data. A mixed model ANOVA detected significant within-subject error in 54 of the 171 ILDs (i.e., 32%). A GLM procedure revealed significant between-observer variation in 14 ILDs and significant observer-by-session differences in 13 ILDs. The majority of these differences involved ILDs with type 3 landmarks as endpoints, such as euryon and alare. Unlike type 1 and 2 landmarks which are biologically informative in all directions, type 3 landmarks contain a substantial arbitrary component. Thus, it is not surprising type 3 landmarks displayed significant digitizing error. Given these results, we caution researchers to be mindful of type 3 landmarks measurement discrepancies when selecting landmarks for coordinate data evaluation.