Some subjective validation methods using graphical displays of data

Abstract

Subjectivemethods for operational validity are presented that use graphical displays of histograms, box plots, and behavior graphs. These methods allow the data to be correlated, have any statistical distribution, and be limited in the number of observations. Model data are used for the reference distribution (instead of a theoretical distribution such as the t or F) and for reference to compare the system data against. These methods are very general and can be used in validating different types of models.

Model reuse versus model development: Effects on credibility and learning

“…Results of 10 replications are automatically exported to Microsoft Excel for review and scenario comparison. Objective validation criteria are provided by a confidence interval for the performance target and subjective validation is operationalised by conceptual model review (Law 2007) and histogram comparison of simulation output (Sargent 1996).…”

Discovering metamodels’ quality-of-fit for simulation via graphical techniques

“…Graphical techniques that are used in the literature in metamodel validation include the use of histograms, box plots, and scatter plots, e.g., see [11][12][13][14][15]. Three new plot types are presented in this work, as discussed in this subsection.…”

Statistical Analyses of Scatterplots to Identify Important Factors in Large-Scale Simulations

“…Two measures of dispersion will be considered: the variance a; of y, and the interquartile range yo75 -yo.25 of y, where yo,75 and yo,25 represent the 0.75 and 0.25 quantiles of y. The variance is the best known measure of dispersion, and the interquantile range is widely used as a summary of dispersion in box plots (e.g., Helton et al, 1996;Sargent, 1996). The interquartile range is less sensitive to outliers than the variance, as is the case in the analogous situation for medians and means.…”