Anthropometric Data Analytics: A Portuguese Case Study

Large amounts of information are systematically generated throughout the course of scientific research and progress. In our case, observations representing the Portuguese population within the central-southern region of Portugal were collected throughout various foetal autopsy procedures. Gestational age (GA) and measured distances and weights of numerous anthropometric features and organs, respectively, were recorded per singleton (24 variables in total). This work seeks to elaborate on the accuracy of different foetal parameters in terms of GA estimation, making use of principal component analysis (PCA) and regression techniques. We created a dataset of 450 foetuses, ranging from 13 to 42 weeks of age, to compute both PCA and regression models. Initial exploratory analysis shed light onto which variables are most explanatory in terms of foetal development, and are thus most likely suitable for predictive rolls. We produced clusters of models, based on coefficient of determination (R) values, by comparing the squared sum of residuals between models (significance level α = 0.05). Models comprised of linear combinations of different variables exhibited significantly higher values of R (p-value ≤ 0.05) when compared to single variable models. Across all regressions, crown-heel length (CHL), crown-rump length (CRL), and foot length (FL) are constantly present within the cluster of best predictors of gestational age. Depending on the type of regression analysis applied, body weight (Body), hand length (HL) also fall onto the same category.