The 15 most frequently selected covariates from the LAR variable selection executions on the 500 unique, 35 observation training sets constructed from the design matrix created by filtering the full design matrix to enforce a maximum permitted correlation coefficient magnitude between remaining covariates pairs of 0.95.

The second column contains the frequencies with which the selected covariates occurred in the 500 selected models. Accompanying each selected covariate in the final column are the covariates from the full design matrix that had correlation coefficient magnitudes with the covariate in question greater than 0.95 and thus were excluded from the design matrix supplied to the variable selection. Colons denote interaction terms for the two covariate terms which the colon separates. Numeric superscripts denote polynomial terms for the covariate indicated by the acronym. Acronyms are expanded in Table 1.