Predicting euarchontan body mass: A comparison of tarsal and dental variables

Abstract

Among individual facets, the ectal and fibular facets of the astragalus and the calcaneal cuboid facet are the best body mass predictors. Since these facets have primarily concave curvature and scale with positive allometry relative to body mass, it appears that candidate skeletal proxies for body mass can be identified based on their curvature and scaling coefficients.

Predicting body mass of bonobos (Pan paniscus) with human‐based morphometric equations

“…Based on several accuracy metrics, the most reliable BMPE is Equation , a panel equation derived from natural‐log transformed bi‐iliac breadth of skeletally immature humans (Figure 3 and Table 3). Since multivariate BMPEs often have greater accuracy than univariate equations (Gingerich, 1990; Jungers, 1990; Mendoza, Janis, & Palmqvist, 2006; Yapuncich et al, 2015), it is somewhat surprising that the equations which include both stature and bi‐iliac breadth (Equations and ) have lower predictive accuracy than the univariate Equation . The inaccuracies of these equations are likely caused by differences in the relationship between stature and body mass in humans and bonobos (Table 2; Figure 2).…”

“…Third, we calculated the mean percentage prediction error (%MPE) by averaging relative %PE by age and sex. Following other studies using these accuracy metrics (Aiello & Wood, 1994; Dagosto & Terranova, 1992; Delson et al, 2000; Elliott et al, 2015a,b; Yapuncich et al, 2015), we consider both a median %PE less than ±20% or conditions in which the majority of test cases have less than ±20%PE to indicate acceptably low error. Finally, we compared observed and predicted BMs using nonparametric two‐tailed Mann‐Whitney U‐tests (to account for nonnormality of observed BMs) to test for significant differences in the group medians.…”

“…To generate this equation, BIB, ST, and BM values were natural log‐transformed. Adult male and female samples were combined, as the level of sexual dimorphism (calculated as [ln(female BM) – ln(male BM)]*100) is only 13%, less than the 20% threshold suggested by Yapuncich et al (2015) to necessitate separate prediction equations. The precision of this equation was gauged with the correlation coefficient ( r 2 ), mean squared error (MSE), and percent standard error of the estimate (%SEE).…”

“…While several theoretical and statistical issues make body mass prediction with linear regression models a deceptively complex procedure (Ruff, 2007; Smith, 2002; Yapuncich, 2018; Yapuncich, Gladman, & Boyer, 2015), extrapolation beyond the limits of the reference sample is one of the most problematic. In a strict sense, extrapolation violates the statistical assumption that the predicted sample belongs to the same population as the reference sample (Smith, 2009).…”

“…Extrapolation provides an opportunity to evaluate the trade‐offs between accuracy and generality to determine when differences (phylogenetic or otherwise) between the reference and test samples render BMPEs unreliable. We have used intentional extrapolation in previous work to determine if BMPEs developed with primate reference samples remain accurate with non‐primate test samples (Yapuncich et al, 2015) and if morphometric equations based on adult humans work well in skeletally immature humans (Walker et al, 2018).…”