In statistics, the techniques that aim to reduce the number of explanatory variables (factors) used to explain observational outcomes. New variables are constructed as combinations of the original variables with the aim of identifying a simpler model structure.