The properties of McReynolds’ constants were studied by a detailed statistical/chemometric analysis. The electronic structure, geometries and hydrophobicity of the McReynolds’ test compounds (benzene, 1-butanol, 2-pentanone, 1-nitropropane, pyridine, 2-methyl-2-pentanol, 1-iodobutane, 2-octyne, 1,4-dioxane and cis-hidrindane) were calculated at the level of PM3 semiempirical quantum chemical method and empirical formulas. The predominant pattern was revealed using cluster and principal component analyses (CA and PCA). Dependence of McReynolds’ constants on the calculated chemical descriptors was modeled by multiple linear regression (MLR) with stepwise selections, principal component regression (PCR) and partial least-square regression (PLSR). A novel statistical approach was developed for case-and-variable selection using the PCR and PLSR methods for characterizing and modeling the polarity of 25 gas chromatography (GC) stationary phases (phthalates, adipates, sebacates, phosphates, citrates and nitrils). Highest occupied molecular orbital energy, dipole moment, averaged isotropic polarizability and the apolar solvent accessible surface area; and energy of the lowest unoccupied molecular orbital and total solvent accessible surface area were suitable to describe the McReynolds’ constants based on the results obtained using Q 2 and adjusted- Q 2. Six of the 10 test compounds were found to be sufficient for the description of the polarity of the columns studied.