Variable Reduction in Multivariate Systems with High Variability to Build up Linear Models Case: Prediction of Boiling Point in Cyclic Hydrocarbons

Abstract

This work presents the development of a methodology supported by inferential statistical methods to reduce the dimensionality of a universe with high variability and collinearity of interest to a population to build up a simple and robust predictive linear contribution model with the least loss of information. As a case study, the reduction of 129 variables is presented, building an inferential model with just one obtaining a correlation coefficient of 0.9595 and a standard deviation of 9.66.