Download The psychometric properties of a shortened Dutch version of the consequences scale used in the PDF

The psychometric properties of a shortened Dutch version of the consequences scale used in the
Name: The psychometric properties of a shortened Dutch version of the consequences scale used in the
Pages: 15
Year: 2017
Language: English
File Size: 1.93 MB
Downloads: 0
Page 2

indicatedtobethebestmodel,havingbothagoodmodelfitandanacceptablecomposite reliability.Inaddition,constructvaliditywasevaluatedtobeacceptable,withgooddiscrimi nantvalidity,althoughtheconvergentvalidityofthefactormeasuring`socialconsequences' couldbeimproved.Concurrentvaliditywasevaluatedasgood.ConclusionsIndecidingwhichmodelbestrepresentsthedata,itiscrucialthatnotonlythemodelfitis evaluated,buttheimportanceoffactorreliabilityandvalidityissuesisalsotakeninto account.Thetwo factormodel,identifyingpersonalconsequencesandsocialconse quences,wasconcludedtobethebestmodel.ThisshortenedDutchversionoftheCADS (CADS_D)isausefultooltoscreenalcohol relatedconsequencesamongcollegestudents.1.IntroductionTheproblematicuseofalcoholandotherdrugshasbeenaworldwideconcernfordecades. Globally,nationalandinternationalpoliciesandinterventionshavebeeninstalledaimingto tackletheharmfulconsequencesofalcoholanddrugmisuse.Actionisespeciallyneeded withregardtoalcoholmisusesinceitisthemostprevalentpsychoactivesubstanceworldwide .AccordingtotheWorldHealthOrganization,theproblematicuseofalcoholremainsone ofthefivemostimportantcausesofdisease,disabilityanddeathacrosstheglobe.Astag gering5.9%ofalldeathsworldwidearecausedbyharmfulalcoholuse.Indeed,alcoholmisuse hasbeenindisputablyidentifiedasapublichealthproblem.Thisisespeciallytruefor youngpeople,suchasuniversityandcollegestudents,asthetransitionfromhighschoolto universityorcollegeisoftenaccompaniedbyhighlevelsofsubstanceuseandmoreproblem aticalcoholuse. Alcoholuseamongstudentshasbeenstudiedextensivelyinrecentyearsandhasreceived muchmediaattention.Alarge scaleFlemishstudyindicatedthat98%ofuniversity andcollegestudentshaveeverusedalcoholand93%ofthesestudentshaddrunkalcoholin thepast12months.Halfofallthesestudents(49.7%)showedriskcharacteristicsofproblem aticalcoholuse.Moreover,excessiveconsumptionpatternssuchasbingedrinking,have beenreportedasacommonpracticeamongyoungpeople,increasingtheriskof experiencingalcohol relatedconsequences.Severalstudieshaveaddressedtheprob lematicconsequencesofstudents'drinkingbehaviour,suchasacademicproblems,injuries, assaults,drivingundertheinfluenceandsexualassault,notonlyharmingthestu dent,butalsootherpeoplearoundthestudentandwithinsocietyasawhole.Giventhe immenseburdenthatalcoholputsonsocietyintermsofhealth,socialandeconomicoutcomes ,itisvitalthatalcoholresearchisbasedonvalidandreliableinstrumentstomeasure theconsequencesofalcoholmisuse.Inrecentdecades,considerableefforthasbeenputinto developingscalestomeasuretheconsequencesofalcohol(anddrug)useamongyoungpeople .However,reliabilityandvaliditytestingofsomeoftheseinstrumentsislacking. TheCoreAlcoholandDrugSurvey(CADS)wasdevelopedin1990asaself reportinstru menttoassessthenature,scopeandconsequencesofalcoholandotherdruguseamongstcol legestudents.Althoughnumerousstudieshaveusedtheconsequencesscale,a subscaleofthesurvey,littleattentionhasbeenpaidtoitspsychometricproperties. Theinitialdeveloperspresentedthisconsequencesscaleasaunidimensionalconstruct, PsychometricpropertiesoftheCoreAlcoholandDrugSurveyPLOSONE|https://doi.org/10.1371/journal.pone.0187876December7,20172/15bySDB.Thefundershadnoroleinstudydesign, datacollectionandanalysis,decisiontopublish,or preparationofthemanuscript. Competinginterests:Theauthorshavedeclared thatnocompetinginterestsexist. Abbreviations:ADF,AsymptoticDistributionFree; AIC,AkaikeInformationCriterion;CADS,Core AlcoholandDrugSurvey inthisresearchthe abbreviationrefersspecificallytothequestion batteryofnegativeconsequences;CFA, Confirmatoryfactoranalysis;CFI,ComparativeFit Index;df,DegreesofFreedom;DUI,Drivingunder theinfluence;DWI,Drivingwhileintoxicated;FL, Factorloading;MI,ModificationIndices;RMSEA, RootMeanSquareErrorofApproximation;SRMR, StandardizedRootMeanSquareResidual;TLI, Tucker LewisIndex;2,Chi square.


Page 3

withoutextensivelyinvestigatingitsfactorstructure,whileanotherresearchstudyfoundthat thisscalehadatwo factorstructurethatidentifiedpersonalconsequences(suchashavinga hangover)andconsequenceswithothers(suchasgettingintoanargumentorfight). Moreover,thesestudieswereallperformedintheUS,creatingadearthofknowledgeofthe factorstructureoftheCADSinothercontexts. Theprimaryaimofthisresearchstudyis,therefore,toassessthepsychometricproperties ofaDutchversionoftheCADSinalargesampleof19,253Flemishuniversityandcollegestu dents.Asalcoholiscurrentlythemostprevalentpsychoactivesubstanceworldwide,wefocus especiallyonassessingthescalewithregardtoalcoholconsequences.Weexaminedthefactor structureoftheDutchCADSbycomparingtheone andtwo factormodelaspresentedinthe literaturebyusingconfirmatoryfactoranalysisandadaptingthemodelsifnecessary. Inaddition,wetestedforcompositereliabilityandbothconstruct(i.e.,convergentanddis criminant)andcriterion related(i.e.,concurrent)validitytoverifytheconsistencyaswellas theaccuracyofthefactors.2.Materialsandmethods 2.1ProcedureandparticipantsTheanalysesarebasedondatacollectedbytheinter universityproject`Headintheclouds' .Across sectionalsurveywassenttostudentsoftheelevenuniversitiesandcollegesin Flanders(Belgium)whowerewillingtoparticipate.Studentswereinvitedbyemailandother methods(e.g.,studentmagazine)toparticipateanonymouslytoanonlinesurvey.Theyhad fourtosixweekstoparticipateintheperiodFebruarytoApril2013andnoreminderwassent .Studentscouldvoluntarilydecidewhetherornottoparticipatebyactivelyclickingonthe linkintheemailwhichwouldleadthemtotheonlinesurvey.Theintroductionclearlystated thatthedatawouldbeanonymous.Toincreaseresponserate,someincentives(e.g.,thechance towinanumberofprizes,includinganiPad)wereofferedtotheparticipatingstudentsand onlyiftheyagreedtoprovideanemailaddress.Fivecollegeswereexcludedfromthesample becausetheyhadaverylowresponserate(<3.5%).Thisresultedinafinalsampleof19,253 collegeanduniversitystudents(22.1%responserate).Thestudywasperformedaccordingto theethicalstandardsoftheAmericanPsychologicalAssociationandwasapprovedbytheEth icsCommitteeofGhentUniversityHospital(ECUZG2013/065). Ofthe19,253participants,35.7%(n=6,867)weremaleand64.3%(n=12,386)were female.Meanagewas21.12years(SD=3.251).Table1providesanoverviewofsexandage distributionsamongparticipatinginstitutions.Wealsoperformedbivariateanalyses(ANO VAs)withageasthedependentvariableandsexaswellasinstitutionasthegroupvariableto verifyanysignificantdifferencesinparticipants'agebetweenmenandwomenandbetween institutions.Resultsindicatedthatagesignificantlydiffersbetweeninstitutions(F(5)=49.733, p<0.000).Withregardtosex,however,nosignificantdifferencewasfoundbetweentheage ofmaleandfemaleparticipants(F(1)=0.117,p=0.732).2.2MeasuresNegativeconsequencesofalcoholuseweremeasuredusingtheCADS.Participantswere askedhowoftentheyhaveexperiencedalistof19consequences(e.g.,gotintoanargumentor fight)asaconsequenceoftheirdrinkingordruguseduringthelastyear.Theinternalconsis tencyoftheitemswasreportedtobehighwithaCronbach'sAlphaof0.90.TheCADS wastranslatedintoDutchbytwoindependenttranslators.Bothtranslationswerealmostsimi lar.Anydifferencesthatdoexistedwerediscussedintheworkinggroupresponsibleforthe questionnaire.Moreover,fivestudentspre testedtheusabilityandcomprehensibilityofthe PsychometricpropertiesoftheCoreAlcoholandDrugSurveyPLOSONE|https://doi.org/10.1371/journal.pone.0187876December7,20173/15


Page 4

questionnaire.TheanswercategoriesoftheCADSwere`none',`one',`two',`threetofive',`six tonine'and`10ormoretimes'.Frequencieswerecodedusingmid pointsofthecategories, respectively0,1,2,4,7.5and11.25timesfortheuppercategory(10timesplushalfrangeto midpointofadjacentcategory).ThecompletelistofconsequencesispresentedinS1 Table . TheAlcoholUseDisorderIdentificationTest(AUDIT)wasdevelopedbytheWorldHealth Organization(WHO)andmeasuresproblematicalcoholusewith10items.Thescalehas proventobeusefulandreliableinmeasuringproblematicalcoholuseamongstudents.TheAUDITwasofficiallytranslatedintoDutchwiththeapprovaloftheWHOand hasproventobeareliablescreeninginstrument.Inthisstudy,weusedtheshortenedver sion,theAUDIT c,whichhasproventobeanequallygoodorevenbetterindicatorformea suringproblematicalcoholuse.TheAUDIT cconsistsofthreequestions,measured ona5 pointscale:`Howoftendoyoudrinkalcohol(ingeneral)';`ifyoudrink,howmany glassesdoyouusuallydrinkperday';`howoftendoesithappenthatyoudrinksixglassesor moreinonesingleoccasion'.ThereliabilityoftheAUDIT cinthepresentstudywasgood ( =0.795). Bingedrinkingwasmeasuredbyaskingstudentstoindicatehowoftentheydrankfour glassesormore(forwomen)orsixglassesormore(formen)duringatimespanoftwohours. Oneglassreferstoastandardglassofalcoholcontaining10gor12.7mlpurealcohol.This amountcorrespondstoapproximately1glassofbeer(25cl),wine(10cl),non distilledbever agesuchassherry(5cl),orspirits(3.5cl).Responseoptionsrangedfrom1=never, 2=lessthanmonthly,3=monthly,4=weekly,to5=dailyoralmostdaily.Thetime frame usedtomeasurebingedrinkingwaswithinthepreviousyear.2.3AnalyticstrategyDatawereanalyzedusingIBMSPSSStatistics22andIBMSPSSAmos22.Weonlyincluded thoseparticipantswhoreporteddrinkingalcoholwithinthepast12months(n=17,756)in theanalyses.Firstly,weperformeddescriptiveanalysestodescribedrinkingcharacteristics andtherelatedalcoholconsequencesinoursample.Next,weexaminedthefactorstructureof theCADSbyperformingconfirmatoryfactoranalyses.Theanalysesareamixofthealterna tivemodelsapproachandamodelgeneratingapproach,asdefinedbyJo reskog,inwhichTable1.Sexandagedistributionsamongparticipatinginstitutions. Sex(%(n))Age(mean(SD)) MaleFemale UniversityofAntwerp (n=1,897)32.4(614)67.6(1,283)22.05(4.273) UniversityofGhent (n=7,181)37.8(2,711)62.2(4,470)21.07(2.182) UniversityofLeuven (n=5,189)33.0(1,713)67.0(3,476)20.80(3.046) KdGCollege (n=2,248)32.3(725)67.7(1,523)21.20(3.825) KHLimburgCollege (n=2,087)32.1(669)67.9(1,418)20.93(3.425) GroupTCollege (n=651)66.8(435)33.2(216)21.84(2.487) Total (n=19,253)35.7(6,867)64.3(12,386)21.12(3.251) https://doi.org/10.1371/journal.pone.0187876.t001 PsychometricpropertiesoftheCoreAlcoholandDrugSurveyPLOSONE|https://doi.org/10.1371/journal.pone.0187876December7,20174/15


Page 5

wecomparetwomodelsaspresentedintheliteratureandmodifythemwiththegoaloffinding amodelthatfitsthedatawellandhasatheoreticallymeaningfulinterpretation.Westarted withtheone factormodelasdescribedbyPresley(i.e.,Model1a)andthetwo factor modelindicatedbyMartensetal.(i.e.,Model2a).Theseinitialmodelswereadaptedand compared,basedonmodelfitandtheircompositereliability.Martensetal.(2005)madesev eraldecisionsintheiranalysestoimprovemodelfit.Firstofalltheyexcludedalltheitems experiencedby5%orlessoftheparticipants.Inaddition,theyexcludeditems11(hadamem oryloss)and12(donesomethingIlaterregretted)astheyloadedhighonbothfactors.We employedasimilarstrategyforourdata. Weusedseveralgoodness of fitindicestomeasuremodelfit.Theclassicgoodness of fit indexis2.However,itiswellknownthat2isalmostalwayssignificantinthecaseoflarge samplesizes.WethereforealsoreportedtheRootMeanSquareErrorofApproximation (RMSEA),StandardizedRootMeanSquareResidual(SRMR),ComparativeFitIndex(CFI) andTucker LewisIndex(TLI).WealsoreportedtheAkaikeInformationCriterion(AIC),as thisindexallowsacomparisonbetweennon nestedmodels.Thefollowing(strict)cutoffcrite riawereusedtoevaluatemodelfit:SRMR<0.08;RMSEA<0.08=adequatefit;< 0.05=goodmodelfit;CFIandTLI>0.95;factorloading(FL)>0.50. SincetheCADSisnotnormallydistributed(0isveryfrequentlyanswered),weusedthe ADFestimatorinAMOS.Item1isthereferenceitemintheone factormodel.Inthe two factormodel,item1isthereferenceitemforthe`personalconsequences'factoranditem 19isthereferenceitemforthefactor`socialconsequences'.WeusedJo reskogRho=(Sum (FL))2/((Sum(FL))2+Sum(1 FL2))toevaluatethecompositereliabilityofeverymodel. Wealsotestedthevalidityofthebestfittingmodel.AsindicatedbytheInternationalTest Commission,weprovidedevidenceonbothconstructvalidityaswellascriterion related validity.Firstofall,constructvaliditywasmeasuredbybothconvergentanddiscriminant validity.AsBrown(2006)describes, convergentvalidityisindicatedbyevidencethatdifferent indicatorsoftheoreticallysimilaroroverlappingconstructsarestronglyinterrelated".In otherwords,allitemsofoneconstructneedtobeinterrelatedwithfactorloadingsabove0.50 (orevenbetterabove0.70).Amorestrictevaluatingtoolofconvergentvalidityismeasuring averagevarianceextracted(AVE=(SumofFL2)/(SumofFL2+Sum(1 FL2)).Strictlyspeaking theAVEneedstobehigherthan0.50.Discriminantvalidity"isindicatedbyresultsshow ingthatindicatorsoftheoreticallydistinctconstructsarenothighlyintercorrelated".Inother words,wedonotwantitemsofoneconstructloadingontoanotherconstruct,oritemsofdif ferentconstructscorrelatingwitheachother.Thecovarianceoffactorsneedstobelowerthan 0.80 0.85.Secondly,wealsoaddressedtheconcurrentvaliditybyreplicatingawell knowncorrelationwithtwoexternalvariables(bingedrinkingandAUDIT c).Missingitems weredeletedusinglistwisedeletion.3.Results 3.1DescriptiveresultsTable2providesthesampleresponsesonbingedrinkingandontheAUDIT c.Table3gives anoverviewoftheitemscoredistributionoftheCADS.3.2Fitoftheone factormodelsModel1a.Westartedbytestingtheone factorstructureoftheCADS,containingallof the19items.AsshowninTables4and5,thefitofmodel1awasbad,exceptfortheRMSEA. 14ofthe19factorloadingswerebelow0.50,andthefactorloadingsofitems13,14and18 werenotsignificantonap<0.001level.Compositereliabilitywasgoodwithrho=0.710. PsychometricpropertiesoftheCoreAlcoholandDrugSurveyPLOSONE|https://doi.org/10.1371/journal.pone.0187876December7,20175/15


Page 6

Model1b.Weexcludedcertainitemsastheywererarelyendorsed(i.e.,5%orless)bythe participants.Thisresultedinanexclusionof8items,namelyitems3,4,13,14,15,16,17 and18.The11 itemscalewastestedasaone factormodel.AsshowninTables4and5,the modelfitwasnotgood.RMSEAindicatedagoodmodelfit,buttheotherfitindicesclearly didnot.Somefactorloadingswerestilllow,althoughallloadingsweresignificant.Compositereliabilitywasadequatewith rho=0.784. Model1c.Becausemodel1bdidnothaveanacceptablefit,weeliminatedonebyoneall theitemswithalowfactorloading(standardizedloading<0.50)fromouranalyses.After eachelimination,weevaluatedthemodelfit,resultinginaone factorstructurecontaining5 items(1,6,8,11,12).Standardizedfactorloadingswerehigherthan0.50andallwerehighly significant(p<0.001).Thismodelwasseentobethe`best'modelofalltheone factormodels. AsshowninTable5,themodelhadanacceptablemodelfit,althoughCFI,andespeciallyTLI couldbeimproved.Compositereliabilitywascalculatedtobe0.812. Fig1presentstheone factormodels.Table2.Drinkingcharacteristicsofthesample. Bingedrinking(%(n))Never38.4 (6612) Lessthan monthly37.7 (6487) Monthly15.5 (2665) Weekly8.1(1387) (Almost)daily0.3(48) AUDIT c(%(n)) AUDIT1 `Howoftendoyoudrinkalcohol(ingeneral)'?Never1.4(242) Monthlyorless23.5 (4071) Onceaweekor less36.9 (6394) 2to3timesa week31.0 (5370) 4timesaweek7.1(1237) AUDIT2 `Ifyoudrink,howmanyglassesdoyouusuallydrinkperday'?1or242.5 (7286) 3or433.1 (5666) 5or614.6 (2507) 7to96.6(1128) 10timesormore3.2(546) AUDIT3 `Howoftendoesithappenthatyoudrinksixglassesormorein onesingleoccasion'?Never25.9 (4441) Lessthan monthly34.6 (5947) Monthly21.4 (3682) Weekly17.2 (2955) (Almost)daily0.9(149) https://doi.org/10.1371/journal.pone.0187876.t002 PsychometricpropertiesoftheCoreAlcoholandDrugSurveyPLOSONE|https://doi.org/10.1371/journal.pone.0187876December7,20176/15


Page 7

3.3Fitofthetwo factormodelsModel2a.Wefirsttestedthetwo factormodelasdescribedinMartensetal.(2005),iden tifyingpersonalconsequences(items1,6,7and8)andconsequenceswithothers(items2,3,4, 5,9,10,19)whichwefurtherrefertoassocialconsequences.WeusedCFAwithcorrelated factors(similartoanobliquerotation)totestthismodel.TheresultsarepresentedinTables4 and5.Allloadingsweresignificant,butnotallofthemwerehigherthan0.50.Moreover,the modelfitwasnotgood(2=407.528;RMSEA=0.023;SRMR=0.088;CFI=0.916;TLI= 0.893;AIC=453.528).Correlationofthetwofactorswas0.82.Compositereliabilitywasgood forfactor1(0.711),butnotforfactor2(0.592). However,asourdatasetisdifferentfromthatofMartensetal.(2005),weextrapolatedthe decisionstheymade(cfr.2.3Analyticstrategy)toourdatasetandtestedtwoadditionalmod els.Asitems3and4wereexperiencedbylessthan5%oftheparticipants,theseitemswere alsoexcludedinouranalyses(Model2b).Sincewedidnotknowwhetheritems11and12 wouldloadhighonbothfactors,weincludedtheminmodeltesting(Model2c).Inthetwo modelsweeliminateditemsifnecessary. Model2b.WhentestingtheinitialModel2b,weconcludedthatthemodelfitwassimilar toModel2a.Factorloadingsweresignificant,butsomewerereallylow(<0.50).Asaconse quence,theseitemsweredeletedonebyoneandmodelfitwasevaluatedeachtime.Thispro cessoftestingandevaluatingthefitresultedinthefollowingmodel:Personalconsequences (items1,6,8)andSocialconsequences(items5,9,19).Theresultsofthismodelareshownin Table4.Allfactorloadingsweresignificantandhigherthan0.50.Themodelfitwasgood,as canbeseeninTable5(2=174.137;RMSEA=0.036;SRMR=0.034;CFI=0.960;TLI= 0.925;AIC=200.137).Correlationofthetwofactorswas0.76.Compositereliabilitywasgood forfactor1(0.773),butnotforfactor2(0.534).Table3.ItemscoredistributionoftheCADS. CADSITEMSPercentageofanswersineachfrequencycategory NeverOnceTwice3 5times6 9times10timesormoreMeanStandardDeviation 131.215.112.417.69.514.33.413.89 286.56.83.42.40.60.30.311.10 395.63.20.80.30.00.00.070.44 495.03.21.00.60.10.10.090.56 587.18.02.91.50.20.20.240.87 641.026.215.211.93.32.41.552.28 791.44.61.61.20.50.70.241.19 854.212.210.011.44.97.41.973.24 977.011.25.93.71.01.20.591.63 1094.52.51.40.80.30.50.171.00 1169.512.37.75.82.32.40.952.18 1267.317.08.65.11.20.80.731.58 1399.70.20.00.00.00.00.010.24 1499.10.70.10.00.00.00.020.28 1599.80.10.10.00.00.00.010.27 1697.11.40.70.50.10.20.080.63 1796.71.70.60.50.20.30.100.79 1899.50.30.10.10.00.00.010.27 1987.27.73.21.30.30.20.240.90 https://doi.org/10.1371/journal.pone.0187876.t003 PsychometricpropertiesoftheCoreAlcoholandDrugSurveyPLOSONE|https://doi.org/10.1371/journal.pone.0187876December7,20177/15


Page 8

Table4.Factorloadings,significance,andcompositereliabilityofthemodels. Model1a:one factormodel19 itemsModel1b:one factormodel11 itemsModel1c:one factormodel5 itemsModel2a:two factor model(pers.:1,6,7,8; soc.:2,3,4,5,9,10,19)Model2b:two factor model(pers:1,6,8; soc:5,9,19)Model2c:two factor model(pers:1,6,8; soc:5,9,12,19) CADS itemsPersonal cons.Social cons.Personal cons.Social cons.Personal cons.Social cons. 10.840(ref.cat.)0.837(ref.cat.)0.827 (ref.cat.)0.835(ref. cat.)0.830 (ref.cat.)0.821 (ref.cat.) 20.354***0.352***0.435*** 30.245***0.333*** 40.265***0.365*** 50.392***0.396***0.463***0.501***0.491*** 60.669***0.670***0.680***0.678***0.681***0.685*** 70.206***0.204***0.203*** 80.670***0.679***0.672***0.683***0.671***0.685*** 90.404***0.405***0.510***0.552***0.548*** 100.236***0.225***0.300*** 110.607***0.609***0.624*** 120.578***0.582***0.588***0.729*** 130.032* 140.046** 150.019*** 160.154*** 170.082*** 180.024 190.408***0.419***0.489(ref. cat.)0.525(ref. cat.)0.514 (ref.cat.) Compositereliability Jo reskog Rho0.7100.7840.8120.7110.5920.7730.5340.7760.662 Significancelevels *p0.05 **p0.01 ***p0.001 Ref.cat.referstothereferencecategoryasexplainedinsection2.3Analyticstrategy https://doi.org/10.1371/journal.pone.0187876.t004Table5.Goodness of fitindicesofthe6models. Goodness of fitindicesModel1aModel1bModel1cModel2aModel2bModel2c 2798.126594.519234.790407.528174.137202.125 df15244543813 p000000 RMSEA0.0160.0280.0530.0230.0360.030 SRMR0.1750.0970.0400.0880.0340.033 CFI0.8680.8820.9480.9160.9600.956 TLI0.8510.8520.8970.8930.9250.929 AIC874.126638.519254.790453.528200.137232.125 p=significancelevel https://doi.org/10.1371/journal.pone.0187876.t005 PsychometricpropertiesoftheCoreAlcoholandDrugSurveyPLOSONE|https://doi.org/10.1371/journal.pone.0187876December7,20178/15


Page 9

Model2c.Model2cwasbasedonModel2b,butincludeditems11and12aswell.Itwas clearthatitem11`hadamemoryloss'belongedtothefactorofpersonalconsequences.For item12`donesomethingIlaterregretted',however,itwassomewhatunclearwhetheritisa consequencethatonlyrelatestothedrinkerortootherpeopleaswell.Wethereforeperformed twoCFA's:onewhereitem12waspartoffactor1andanotherwhereitbelongedtofactor2. AsthesecondCFAgaveabetterfit(AICof409.557comparedto385.866),weincludeditem 12inthefactorofsocialconsequences. However,sinceitem11hadahighcrossloading(similartoMartensetal.(2005))withthe factorsocialconsequences,westillexcludeditem11fromthemodel.Thisresultedinamajor improvementofthemodelfit.Thisprocessoftestingandevaluatingfitresultedinthefollow ingmodel:Personalconsequences(items1,6,8)andSocialconsequences(items5,9,12,19). TheresultsofthismodelareshowninTables4and5.Factorloadingswereallsignificantand modelfitwasgood(2=202.125;RMSEA=0.030;SRMR=0.033;CFI=0.956;TLI=0.929; AIC=232.125).Correlationbetweenthetwofactorswas0.78andcompositereliabilityforthe twofactorswas0.776and0.662. Fig2presentsthetwo factormodels. Fig1.Overviewofthethreeone factormodels. https://doi.org/10.1371/journal.pone.0187876.g001 PsychometricpropertiesoftheCoreAlcoholandDrugSurveyPLOSONE|https://doi.org/10.1371/journal.pone.0187876December7,20179/15


Page 10

3.4Conclusion`best'modelWeperformedCFA'sonbothone andtwo factorstructuresoftheCADS,startingfromtwo modelsintheliterature,andadaptingthembasedonthe(significanceof)factorload ings,modificationindicesandgoodness of fitindices.AsshowninTable5,Model2bhasthe lowestAICandthusthebestmodelfit,closelyfollowedbyModel2c.However,thefactor `socialconsequences'ofModel2bhasalowcompositereliability,whichismuchbetterin Model2c.Indecidingwhichmodelbestrepresentsthedata,itiscrucialthatnotonlymodelfit isevaluated,butalsocompositereliabilityistakenintoaccount.Basedonthis,itcanbecon cludedthatModel2cisthebestfittedmodelinunderstandingtheconsequencesofalcohol misuse,asithasbothagoodmodelfitandanacceptablecompositereliability.Inthenextstep, wetesttheconstructandconcurrentvalidityofthismodel.3.5ConstructandconcurrentvalidityWeevaluatedthevalidityofModel2candwefocusedonbothconstructandconcurrent validity. 3.5.1Constructvalidity.Constructvaliditywasmeasuredbybothconvergentanddis criminantvalidity.Asdescribedintheanalyticstrategy(section2.3),allitemsofaconstruct needtobehighlyinterrelated(factorloadings>0.50)tomeasureconvergentvalidity.Model 2ccomplieswiththisstandard,andinparticularthefactorloadingsofpersonalconsequences areveryhigh.Onlyitem5hasaslightlylowerfactorloading(0.49).Thestricterevaluatingtool ofconvergentvalidity(AVE),however,showsmixedresults.Factor1withanAVEof0.537 hasagoodconvergentvalidity.Factor2,withanAVEof0.334,however,hasalowerconver gentvalidity.IftheAVEis<0.50,thismeansthatthevarianceofthemeasurementerroris largerthanthevarianceexplainedbythefactor,whichmakesthevalidityofthefactorandthe individualindicatorsquestionable.Thevalidityoffactor2isthuslessstrongthanthatof factor1.Nevertheless,allfactorloadingsaresignificantandclosetoorlargerthan0.50.Fur thermore,thefactorshaveahighdiscriminantvalidity,astherearenocross loadingsbetween theindicatorsofthetwofactorsandthecovarianceofthetwofactorsislowerthanthethresh oldof0.80 0.85. Fig2.Overviewofthethreetwo factormodels. https://doi.org/10.1371/journal.pone.0187876.g002 PsychometricpropertiesoftheCoreAlcoholandDrugSurveyPLOSONE|https://doi.org/10.1371/journal.pone.0187876December7,201710/15


Tags: Download The psychometric properties of a shortened Dutch version of the consequences scale used in the PDF, The psychometric properties of a shortened Dutch version of the consequences scale used in the free pdf download, The psychometric properties of a shortened Dutch version of the consequences scale used in the Pdf online download, The psychometric properties of a shortened Dutch version of the consequences scale used in the 2017 download, The psychometric properties of a shortened Dutch version of the consequences scale used in the.pdf, The psychometric properties of a shortened Dutch version of the consequences scale used in the read online.
About | Contact | DMCA | Terms | Privacy | Mobile Specifications
Copyright 2021 FilePdf