Comparing Simulated and Measured Values Using Mean Squared Deviation and its Components

• DOI: 10.2134/agronj2000.922345x
• Publication Date: 2010-07-28T19:00:05Z
• Authors: Kazuhiko Kobayashi; Moin Us Salam
• Keywords: Absolute deviation; Root mean square

When output ( x ) of a mechanistic model is compared with measurement y ), it common practice to calculate the correlation coefficient between and , regress on . There are, however, problems in this approach. The assumption regression, that linearly related not guaranteed unnecessary for – comparison. regression coefficients are explicitly other commonly used statistics [e.g., root mean squared deviation (RMSD)]. We present an approach based (MSD = RMSD 2 show better suited comparison than regression. Mean sum three components: bias (SB), difference standard deviations (SDSD), lack weighted by (LCS). To examples, MSD‐based analysis was applied simulation vs. comparisons literature, results were those from analysis. MSD clearly identified contrasts larger others; correlation–regression tended focus lower line far equality line. It also shown easier interpret This because components simply additive all constituents explicit. will be useful quantify calculated values obtained measurements.