Abstract
The estimation from long bone lengths of stature in humans or body size in apes has a deep history in physical anthropology. To date, we can enumerate at least five different statistical methods for making such estimations. These methods are: (1) the regression of body length on long bone length (inverse calibration), (2) regression of long bone length on body length followed by solving for body length (classical calibration), (3) major axis regression of body length on long bone length, (4) reduced major axis regression of body length on long bone length, and (5) use of a long bone/body length ratio. We examine some of the statistical properties of these estimators using a large sample of humans (n=2053) to derive the estimators, and applying them to smaller samples ofPan troglodytes(n=42),Pan paniscus(n=8), andGorilla gorilla(n=35).
Based on the root mean-squared error (RMSE), the reduced major axis is the preferred estimator for body length in the combinedPansample. However, inverse calibration is the best estimator for body length in gorillas based on the RMSE. Many estimators grossly underestimate body length in the apes. Differences in allometries between humans and great apes are obvious, but it is important to show the assumptions necessary in estimating body size from fossil remains, especially when isolated long bones are recovered and the global allometry is consequently unknown.