Research Design for Simplifying Anthropometric Data Collection Process Using PCA

Abstract

The current study is aiming at demonstrating how to reduce the numbers of variables using Principal Component Analysis (PCA) to be measured in an anthropometry survey that involves sample data of large number of variables. A case study of analysis of anthropometric data involving PCA method has been reported based on the anthropometric data (32 anthropometric variables) of Ethiopian army personnel collected from 250 male participants. The linear regression models were also constructed using least square method to predict the regression equation of relevant body variables that had correlation coefficients (R) > 0.70. Variables having the lesser factors loading coefficient (<60%), commonality and correlation coefficients (<70%) were counted as independent variables and included in a minimum data set for measurement. The PCA provided six principal component factors. Total 20 regression equations of dependent variables were constructed from the six influential predictors. Therefore, we observed that the total 12 variables (six dominant variables, five variables with less commonality and/or correlation coefficient from their respective predictors, and one targeted variable ‘mass’) can create a minimum data set that almost accounts for the variability produced by 32 original variables. Hence, during data collection, 12 independent anthropometric variables can extensively represent 32 variables. The current case study would help the researchers to save time and reduce their anthropometric survey to these 12 variables that can predict the remaining ones. It would also guide the researchers to adopt PCA to identify the representative anthropometric variables from large number of variables.