Notes. “Best” indicates the entire galaxy datasets with molecular gas masses calculated using a variable α_CO, for which results are shown in Fig. 4; for S∕N > 3, only the detections are considered; “Const. α_CO” represents the whole galaxy sample, where a constant α_CO is used to convert CO luminosities into molecular gas masses; “Only APEX” marks the fit results when only the APEX data are used. In the columns, m and q indicate slope and intercept of the relations, respectively, inferred from PCA; ⟨ΔSFMS_g⟩, ⟨log (SFE_b)⟩, and ⟨log(f_mol,b)⟩ show the medians of global ΔSFMS, beam SFE, and beam f_mol of the distributions, respectively; r_p and r_s are the Pearson and Spearman correlation coefficients, respectively. For most of the correlation realizations we obtain p-values largely below 10⁻⁵ from both correlation tests, except for ΔSFMS_g-SFE_b relations for the centrally star-forming galaxies, for which we measure Pearson p-values of the order of 10⁻² and Spearman p-values of the order of 10⁻¹. The uncertainties are obtained by 1000 bootstrap iterations of the PCA fit, and are provided as 75th –50th percentiles and 50th –25th percentiles of the m and q distributions.