Page 1
• RajinderKoul,Ph.D.,CCC‐SLP• ProfessorandChairperson• DepartmentofSpeech,LanguageandHearingSciences• TexasTechUniversityHealthSciencesCenter
EXPERIMENTAL AND NON-EXPERIMENTAL RESEARCH IN BEHAVIORAL SCIENCES: IMPLICATIONS FOR PRACTICE
UNIVERSITY OF PRETORIACENTRE FOR AUGMENTATIVE AND ALTERNATIVE COMMUNICATION (CAAC)
Page 2
RESEARCH DESIGNResearchdesignsareastructureoftemporal‐spatialarrangementswithinwhichtheselectedvariablesarecontrolled,manipulated,andmeasured(Hegde,2003).
Researchdesignshelprevealrelationshipsbetweenandamongvariables: Independent:notrelated Cause‐effectrelationships Correlativerelationships Interactiverelationships:multiplevariablescombinetogethertoproduceaneffect;thetotaloutcomeistypicallygreaterthanthemereadditiveeffects
Page 3
NON‐EXPERIMENTAL RESEARCH
Lackofcontroloverextraneousand/orassignedindependentvariables(IVs)
Independentvariablesarenotintroduced,withdrawn,ormanipulated
Relationshipbetweencauseandeffectcannotbeestablished
Page 4
EX‐POST FACTO RESEARCH
Page 5
EX‐POST FACTO
Ex‐PostFactoResearch(afterthefact,“retrospectivesearch”) Independentvariableshaveoccurredinthepast‐ oneisstudyingthedependentvariable(DVs)
Unabletodeducecause‐effectrelationshipsbecauseindependentvariablescannotbemanipulated
Casehistorymethod
Page 6
EX‐POST FACTO
Weaknesses ItisnotexperimentalbecauseIVsarenotmanipulated
LackscontrolbecauseextraneousIVsarenotruledout(internalvalidityconcerns)
CannotsupporttheoryStatisticalcorrelations
Page 7
EX‐POST FACTO
AdvantagesUncoversimportantvariablesExploratoryscienceSocialvs.scientificviewpoint
Page 8
NORMATIVE RESEARCH
Page 9
NORMATIVE RESEARCH
AlsocalleddevelopmentalordescriptiveresearchDistributionofselecteddependentvariablesacrossdifferentagegroupsisobservedandrecorded
Helpsdifferentiatetypicalfromatypical(standardizedtests)
HelpsguideclinicalinterventionNotabletomanipulateIndependentvariablesandthuscannotdrawcause‐effectrelationship
Page 10
NORMATIVE RESEARCH
Howtoconduct: IdentifypopulationDrawarepresentativesample(i.e.,allparticipantshaveanequalchanceofparticipatinginthisstudy)
SamplethebehaviorofselectedparticipantsGeneralizeresultstothepopulation
Page 11
NORMATIVE RESEARCH
WeaknessesDifficultyobtainingtruerandomsamples Inadequate(brief)samplingofresponsesDoesnotsupportatheory
TwoTypesCrossSectionalresearchLongitudinalresearch
Page 12
STANDARD‐GROUP COMPARISONS
Page 13
STANDARD‐GROUP COMPARISON
Extensionofnormativeresearch GroupsareformedonthebasisofoneDVand
arethencomparedonthebasisofthesameoranotherDV
TwostrategiestoperformStandard‐GroupComparison:
1. Criterionandcomparisonmeasuresarethesame2. Criterionandcomparisonmeasuresaredifferent
Page 14
STANDARD‐GROUP COMPARISON
Strengths:CandescribedifferencesHelpsdistinguishclinicalgroups
Weaknesses:Cannotassigncause/effectrelationshipsCannotmanipulateIV
CohortStudiesandCase‐controlstudiesarevariationsofStandard‐groupcomparisonresearch
Page 15
SURVEY RESEARCH
Assesssomefeaturesofagroupofpeopleoraparticularsociety
DistributionofDVinthepopulationisdetermined(+or‐)4‐6%rangeoferror
Errorratedependsupontherepresentativenatureofthesample
Realthoughtsandactionsmaynotbethesameasopinionsandattitudes
Page 16
EXPERIMENTAL RESEARCH
Page 17
EXPERIMENTAL RESEARCH
ControlledarrangementinwhichoneormoreIVsaremanipulatedandtheeffectsofsuchmanipulationontheDVsaremeasuredCanisolatecause/effectrelationship(seekstoexplainaneventbydiscoveringitscausalvariables)
Control‐canruleoutcompetinghypothesisCanexplainphenomenaandsupporttheories
Page 18
EXPERIMENTAL RESEARCH
Twovarieties:BasicandAppliedResearchClinical:usedtodescriberesearchthatisinsomewayconnectedwithdiseasesanddisorders;researchaimedatunderstandingandtreatingvariousdisordersofcommunicationisclinical
Applied:non‐clinicalasinbuildingbetterbridgesorsaferhighways
Page 19
EXPERIMENTAL RESEARCH
BasicResearch:Givesusinsightstodoappliedresearch‐veryimportantOftentheoretical
Page 20
EXPERIMENTAL RESEARCH
Twostrategiestoruleoutextraneousvariablesa) Groupstrategyb) Singlesubjectstrategy
Page 21
EXPERIMENTAL RESEARCH‐BETWEEN GROUP STRATEGYIngroupdesigns,youcomparemeansandstandarddeviationsof2ormoregroupsondependentvariablesofinterest(e.g.,randomizedcontroltrials)
Useinferentialstatisticstoseeifthedifferencesacrossthegroupsaresignificant.
Taketheaverageperformanceandcompareacrossgroups,thendetermineifthegroupsaredifferent.
Page 22
EXPERIMENTAL RESEARCH‐BETWEEN GROUP STRATEGY
Randomization:randomselectionandassignment
Matching:MatchingpairsofparticipantsEquivalentfrequencydistribution‐matchedonastatisticalbasisGroupsareequalonassignedvariablesexceptontheIVthatoneismanipulated
Page 23
EXPERIMENTAL RESEARCH‐BETWEEN GROUP STRATEGY
Wellcontrolledgroupdesignshaveatleasttwogroups(onecontrolandoneexperimental)
Majorityofgroupdesignsmeasurepre/posttestdifferences
Groupdesignshaveinferentialgenerality(sampletothepopulation),notlogicalgenerality
Page 24
EXPERIMENTAL RESEARCH‐BETWEEN GROUP STRATEGY
Typesofgroupdesigns:Pre‐ExperimentalDesignsTrueExperimentalDesignsQuasi‐experimentaldesignsCorrelationalDesigns
Page 25
TREATMENT RESEARCH
Uncontrolledtreatmentresearch (e.g.,casestudies)Controlledtreatmentresearch‐establishacause/effectrelationshipbetweentreatmenttechniqueandtheDV(i.e.,positivechangesinthepatientswhoreceiveit)
Directlyreplicatedtreatmentresearch‐sameinvestigators,originalsetting,differentsubjects(samemethodology)
Systematicallyreplicatedtreatmentresearch‐othercliniciansinothersettings
Page 26
EVIDENCE HIERARCHY
Page 27
QUESTIONS/DISCUSSION