Seleção de variáveis: uma aplicação a dados de moinho de cimento

Abstract

Having as object of study a certain mill that produces cement, we use two methods of variable selection, LASSO and stepwise, to identify variables that influence the engine power and, consequently, impact the cement production. We also considered lagged covariates at 4 different times and performed the diagnostic analysis for the estimated models with identification of influential points. Among all the analyzed models, we chose the selection that was made by stepwise without influential points and without time lags, which has the lowest value for the selection criteria, AIC and BIC.