模拟中统计散点的分析

ProceduresforidentifyingpatternsinscatterplotsgeneratedinMonteCarlosensitivityanalysesaredescribedandillustrated.These
proceduresattempttodetectincreasinglycomplexpatternsinscatterplotsandinvolvetheidentificationof(i)linearrelationshipswith
correlationcoefficients,(ii)monotonicrelationshipswithrankcorrelationcoefficients,(iii)trendsincentraltendencyasdefinedbymeans,
mediansandtheKruskal–Wallisstatistic,(iv)trendsinvariabilityasdefinedbyvariancesandinterquartileranges,and(v)deviationsfrom
randomnessasdefinedbythechi-squarestatistic.Asequenceofexampleanalyseswithalargemodelfortwo-phasefluidflowillustrateshow
theindividualprocedurescandifferinthevariablesthattheyidentifyashavingeffectsonparticularmodeloutcomes.Theexampleanalyses
indicatethattheuseofasequenceofproceduresisagoodanalysisstrategyandprovidessomeassurancethatanimportanteffectisnot
overlooked.