Projection of Anthropometric Correlation for Virtual Population Modelling

A new statistical method for generation of virtual populations based on anthropometric parameters is developed. The method addresses the problem that most anthropometric information is reported in terms of summary data such as means and standard deviations only, while the underlying raw data, and therefore the correlations between parameters, are not accessible. This problem is solved by projecting correlation from a data set for which raw data are provided. The method is tested and validated by generation of pseudo females from males in the ANSUR anthropometric dataset. Results show that the statistical congruency of the pseudo population with an actual female population is more than 90% for more than 90% of the possible parameter pairs. The method represents a new opportunity to generate virtual populations for specific geographic regions and ethnicities based on summary data only.