Ten measurements were taken with each instrument at each ethanol reference vapor concentration, as well as ten measurements of each potentially interfering substance concentration. Measurements were alternated with the Pro, C8, and C6 at approximately 2-minute intervals. The mean, percent coefficient of variation (%CV), percent bias, R-squared, and ordinary least squares regression was calculated. Apparent ethanol response to potential interfering substances was recorded.
An evaluation of the potential sources of uncertainty was evaluated using an Ishikawa diagram (Ishikawa, 1989, pp. 229–232) seen in Figure 1 . Standard uncertainties were combined using the root-sum-squares approach where,\(\mathrm{u}_{\mathrm{\text{combined}}}\mathrm{=}\sqrt{{{\mathrm{u}_{\mathrm{\text{reference}}}^{\mathrm{2}}\mathrm{+u}}_{\mathrm{\text{precision}}}^{\mathrm{2}}\mathrm{+u}}_{\mathrm{\text{bias}}}^{\mathrm{2}}}\), as detailed in the literature (Anghel, 2008; Archer, De Vos, & Visser, 2007; Brockley-Drinkman & Barkholtz, 2019; Rod G. Gullberg, 2006; Hwang, Beltran, Rogers, Barlow, & Razatos, 2016; Hwang, Rogers, Beltran, Razatos, & Avery, 2016; Philipp et al., 2010; Souza et al., 2006). Bias was incorporated as a component in the estimate of measurement uncertainty, expanding the coverage interval (Magnusson & Ellison, 2008; Phillips, Eberhardt, & Parry, 1997). The combined standard uncertainty was expanded to the 95% coverage interval (\(U=2u_{\text{combined}}\)). Figure 2 shows the individual uncertainty component’s percent contribution to the combined standard uncertainty.