Variable preparation
Reduction of multicollinearity for all variables was performed by constructing a correlation matrix and performing hierarchical cluster analysis, which groups variables according to their mutually related correlations (Benito, Cayuela, & Albuquerque, 2013; Sarstedt & Mooi, 2014; Albuquerque et al., 2018). A cutoff of 0.5 Pearson’s correlation index was used; all variables correlated higher than 0.5 were discarded (Albuquerque et al.). Biserial correlation analysis, with variables correlated to presence/absence data for Cochemiea halei , was performed for all variables below 0.5 (Kraemer, 2006; Stolar & Nielsen, 2015, Albuquerque et al.). From each cluster of correlated variables as derived from the hierarchical cluster analysis, the variable with the highest correlation to the distribution of C. halei was chosen for use in modeling.