Using Multivariate Analysis within the Vertebral Column to Identify Individual Vertebrae

Abstract

This article demonstrates the utility of a multivariate analysis of vertebrae in an applied context. The human vertebral column is a morphologically complex group of elements. Current methods rely on morphological characteristics to classify isolated vertebrae qualitatively. This research provides a bridge between morphological assumptions for vertebral designations and quantitative classification. These osteometric methods and statistical analyses provide quantifiable information relating to the accuracy of vertebrae classification. The sample used for this analysis consists of osteometric vertebral measurements from intact vertebral columns from 59 individuals. In order to assess the potential for these vertebral measurements to classify vertebrae, regional grouping models based on vertebral column segments were developed and analyzed. The data were tested for multivariate normality and homogeneity of variance–covariance matrices in order to comply with the assumptions required by the statistical analyses used for classification. Linear discriminant function analysis was used for classification. The sensitivity and specificity of each vertebral group prediction were used for evaluation. This research demonstrates that by using osteometric methods and statistical analyses, the accuracy of vertebrae classification is quantifiable. This method has been developed to assist with the sorting and analysis of commingled and fragmentary skeletal remains.