ESTIMATION OF AFRICAN APE BODY LENGTH FROM FEMUR LENGTH

The estimation from long bone lengths of stature in humans or body siz e 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 le ngth on body length followed by solving for body length (classical cal ibration), (3) major axis regression of body length on long bone lengt h, (4) reduced major axis regression of body length on long bone lengt h, and (5) use of a long bone/body length ratio. We examine some of th e statistical properties of these estimators using a large sample of h umans (n=2053) to derive the estimators, and applying them to smaller samples of Pan troglodytes (n=42), Pan paniscus (n=8), and Gorilla gor illa (n=35). Based on the root mean-squared error (RMSE), the reduced major axis is the preferred estimator for body length in the combined Pan sample. However, inverse calibration is the best estimator for bod y length in gorillas based on the RMSE. Many estimators grossly undere stimate body length in the apes. Differences in allometries between hu mans and great apes are obvious, but it is important to show the assum ptions necessary in estimating body size from fossil remains, especial ly when isolated long bones are recovered and the global allometry is consequently unknown. (C) 1998 Academic Press Limited.