engleski

Data archaeology in archaeometry and osteometry: can we simulate raw data from descriptive statistics?

In all scientific fields, the external validity of the studies is an indispensable condition to generalize the results. So, validation and cross- population studies are always of paramount importance. Although researchers are encouraged to provide and publish their raw data, it is still more exception than a rule. Also, many earlier studies contain valuable data, but it is not possible to access them. Recent anthropological studies provided a novel approach that employs algorithms used to recreate data from descriptive statistics. Using the assumption that skeletal metric data from a homogenous sample follow the normal distribution, they provided raw data and constructed metapopulation data. Although that approach showed promising results, it was not tested in detail if the original sample and simulated sample differed. This study aimed to examine differences between the original and simulated sample on seven femoral measurements. For the original sample, we used those of male and female skeletons from Ancient Salona for which we computed descriptive statistics. Using this data and package ‘truncnorm’ in R, we created a simulated sample. We statistically compared distributions of those samples and descriptive data. In both samples, we tested the accuracy of univariate discriminant functions for sex estimation on femoral measurements developed in previous research for various ancient and modern populations. For all variables, both for males and females, there were no statistically significant differences. Also, in most of the cases, discriminant functions for sex estimation for the previous study showed minor variation in accuracy when they were applied to the original and simulated sample. Therefore, the employed approach could be of great importance in anthropological population studies but also studies from the forensic context for validation and development of methods for the estimation of biological profile. It could also be utilized for archaeometric studies of materials whose measurements follow normal distributions.

simulated data ; truncated normal distribution ; sex estimation ; osteometry ; archaeometry ; forensic science

nije evidentirano