Improved leaps and bounds variable selection algorithm based on principal component analysis

Geochemistry of selenium, barium, and iodine in representative soils of the Brazilian Amazon rainforest

“…The following procedure was applied for each response variable: 1) Select explanatory variables to obtain a list of candidates for the multiple linear regression model based on the significance (p ≤ 0.05) and correlation coefficient (|r| ≥ 0.4) of each candidate in the Spearman correlation matrix (Fig. S1 of the supplementary material); 2) For the multiple linear regression model, the raw data was submitted to log-transformation in order to attend parametric statistical assumptions (normality and homoscedasticity) whenever necessary, and the best combination of explanatory variables was obtained with regression by leaps and bounds approach using the branch and bound algorithm (Furnival and Wilson Junior, 2000;Zhang et al, 2014).…”