engleski

Statistical modelling of anthropometric characteristics evaluated on nutritional status

The aim of this chapter is to illustrate to researchers in anthropometry and nutrition sciences how to apply chemometric methods for extraction of data structures and their relations leading to development of mathematical models. Formally, this chapter gives a short review of basic methodologies in multivariate chemomemetric analysis from theoretical and practical view points and illustrates its high potential in analysis of complex life systems. As an example, here is chemometry applied for analysis of anthropometric parameters and nutritional status for secondary school girls and boys of age 14 to 18 situated in boarding schools in Croatia. In total are measured 25 variables ; anthropometric, energy and macronutrients (energy, fat, proteins, and carbohydrates) and some life style factors. The data cover the period from 1997 to 2010 with the total number of 3440 individual recordings. The focus of the analysis is to determine the data structure of normal, overweight and obese subpopulations and relate their classification to principal latent variables and to evaluate corresponding main important factors. Analysis shows clear appearance of body mass index (BMI) bimodal monovariable probability density function and its dynamics through the period of 15 years. Developed are and compared PLS models for body fat prediction based on Carnegie Mellon University data and boarding schools in Croatia. Generally, it is shown that chemometric analysis provides effective tools for data classification and extraction of important variables in analysis of complex phenomena such as nutritional status and obesity.

chemometrics, principal component analysis PCA, linear discriminate analysis LDA, partial least squares PLS, variable importance analysis, bimodal monovariable probability density distribution, nutritional status, anthropometry, secondary boarding school, adolescence, BMI, body fat, obesity

nije evidentirano