Complex Data Modeling and Computationally Intensive Statistical Methods (2010)

Recentyearshaveseentheadventanddevelopmentofmanydevicesabletorecordand storeaneverincreasingamountofinformation. Thefastprogressofthesetechnologies is ubiquitousthroughoutall ?elds of science and applied contexts, ranging from medicine, biologyandlifesciences, toeconomicsandindustry. Thedataprovided bytheseinstrumentshavedifferentforms:2D-3Dimagesgeneratedbydiagnostic medicalscanners, computervisionorsatelliteremotesensing, microarraydataand genesets, integratedclinicalandadministrativedatafrompublichealthdatabases, realtimemonitoringdataofabio-marker, systemcontroldatasets. Allthesedata sharethecommoncharacteristicofbeingcomplexandoftenhighlydimensional. Theanalysisofcomplexandhighlydimensionaldataposesnewchallengesto thestatisticianandrequiresthedevelopmentofnovelmodelsandtechniques, fueling manyfascinatingandfastgrowingresearchareasofmodernstatistics. Anincomplete listincludes for example: functionaldata analysis, that deals with data having a functionalnature, suchascurvesandsurfaces;shapeanalysisofgeometricforms, that relatestoshapematchingandshaperecognition, appliedtocomputationalvisionand medicalimaging;datamining, thatstudiesalgorithmsfortheautomaticextraction ofinformationfromdata, elicitingrulesandpatternsoutofmassivedatasets;risk analysis, fortheevaluationofhealth, environmental, andengineeringrisks;graphical models, thatallowproblemsinvolvinglarge-scalemodelswithmillionsofrandom variableslinkedincomplexwaystobeapproached;reliabilityofcomplexsystems, whoseevaluationrequirestheuseofmanystatisticalandprobabilistictools;optimal designofcomputersimulationstoreplaceexpensiveandtimeconsumingphysical experiments. Thecontributionspublishedinthisvolumearetheresultofaselectionbasedonthe presentations(aboutonehundred)givenattheconference"S. Co. 2009: Complexdata modelingandcomputationallyintensivemethodsforestimationandprediction", held ? atthePolitecnicodiMilano. S. Co. isaforumforthediscussionofnewdevelopments ? September14-16,2009. Thatof2009isitssixthedition, the?rstonebeingheldinVenice in1999. VI Preface andapplicationsofstatisticalmethodsandcomputationaltechniquesforcomplexand highlydimensionaldatasets. Thebookisaddressedtostatisticiansworkingattheforefrontofthestatistical analysisofcomplexandhighlydimensionaldataandoffersawidevarietyofstatistical models, computerintensivemethodsandapplications. Wewishtothankallassociateeditorsandrefereesfortheirvaluablecontributions thatmadethisvolumepossible. MilanandVenice, May2010 PietroMantovan PiercesareSecchi Contents Space-timetextureanalysisinthermalinfraredimagingforclassi?cation ofRaynaud'sPhenomenon GrazianoAretusi, LaraFontanella, LuigiIppolitiandArcangeloMerla. . . . . . 1 Mixed-effectsmodellingofKevlar?brefailuretimesthroughBayesian non-parametrics RaffaeleArgiento, AlessandraGuglielmiandAntonioPievatolo. . . . . . . . . . . . 13 Space?llingandlocallyoptimaldesignsforGaussianUniversalKriging AlessandroBaldiAntogniniandMaroussaZagoraiou. . . . . . . . . . . . . . . . . . . . 27 Exploitation, integrationandstatisticalanalysisofthePublicHealth DatabaseandSTEMIArchiveintheLombardiaregion PietroBarbieri, Niccolo`Grieco, FrancescaIeva, AnnaMariaPaganoniand PiercesareSecchi. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 Bootstrapalgorithmsforvarianceestimationin PSsampling AlessandroBarbieroandFulviaMecatti. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57 FastBayesianfunctionaldataanalysisofbasalbodytemperature JamesM. Ciera. . . . . . . . . . . . . . . . . . . . . . . .