Designation:E2617−10StandardPracticeforValidationofEmpiricallyDerivedMultivariateCalibrations1ThisstandardisissuedunderthefixeddesignationE2617;thenumberimmediatelyfollowingthedesignationindicatestheyearoforiginaladoptionor,inthecaseofrevision,theyearoflastrevision.Anumberinparenthesesindicatestheyearoflastreapproval.Asuperscriptepsilon(´)indicatesaneditorialchangesincethelastrevisionorreapproval.1.Scope1.1Thispracticecoversrequirementsforthevalidationofempiricallyderivedcalibrations(Note1)suchascalibrationsderivedbyMultipleLinearRegression(MLR),PrincipalCom-ponentRegression(PCR),PartialLeastSquares(PLS),Artifi-cialNeuralNetworks(ANN),oranyotherempiricalcalibra-tiontechniquewherebyarelationshipispostulatedbetweenasetofvariablesmeasuredforagivensampleundertestandoneormorephysical,chemical,quality,ormembershippropertiesapplicabletothatsample.NOTE1—Empiricallyderivedcalibrationsaresometimesreferredtoas“models”or“calibrations.”Inthefollowingtext,forconciseness,theterm“calibration”maybeusedinsteadofthefullnameoftheprocedure.1.2Thispracticedoesnotcoverproceduresforestablishingsaidpostulatedrelationship.1.3Thispracticeservesasanoverviewoftechniquesusedtoverifytheapplicabilityofanempiricallyderivedmultivari-atecalibrationtothemeasurementofasampleundertestandtoverifyequivalencebetweenthepropertiescalculatedfromtheempiricallyderivedmultivariatecalibrationandtheresultsofanacceptedreferencemethodofmeasurementtowithincontrollimitsestablishedfortheprespecifiedstatisticalconfi-dencelevel.1.4Thisstandarddoesnotpurporttoaddressallofthesafetyconcerns,ifany,associatedwithitsuse.Itistheresponsibilityoftheuserofthisstandardtoestablishappro-priatesafetyandhealthpracticesanddeterminetheapplica-bilityofregulatorylimitationspriortouse.2.ReferencedDocuments2.1ASTMStandards:2E131TerminologyRelatingtoMolecularSpectroscopyE1655PracticesforInfraredMultivariateQuantitativeAnalysisE1790PracticeforNearInfraredQualitativeAnalysis3.Terminology3.1Forterminologyrelatedtomolecularspectroscop...
