“…Sample sizes for (i.e., number of points in) these calibration curves depended on the number of calibration CRMs for which an analyte concentration was available, one to three (Table II); since three replicates were run on each calibration CRM, three to nine data points were used per calibration curve. 23 Na, 24 Mg, 27 Al, 31 P, 39 K, 44 Ca, 49 Ti, 51 V, 52 Cr, 55 Mn, 57 Fe, 59 Co, 60 Calibration curves were obtained using linear models, forced through the origin, and fitted separately for each analyte (a) in each sample set (s). Specifically, median signal intensity (I a,s ; counts per second) was used as the dependent variable and the corresponding theoretical concentrations (mg/g), C a,s , was set as the independent variable; i.e., I a,s ¼ b a,s C a,s , where b a,s is the calibration curve slope for analyte a in sample set s. Concentrations (mg/g dry weight) of analyte a for test samples in set s were determined (C 0 a,s ) from intensity values as C 0 ,s ¼ I ,s ,s .…”