ExperimentalDesign

1、Experimental DesignTypes - Broad DistinctionQualitativeQuantitativeQualitative ResearchNarrativeSubjects own wordsSummarizes behaviorsDescriptiveMethodsInterviewsObservation notesSurveysQuantitative ResearchNumerical data collected & analyzedExplores relationship between variablesIndependent (single

2、 or multi-leveled)Dependent (single or multiple)Analysis may permit exploration of an interaction between variable Experimental or quasi-experimentalExperimental DesignsCharacterized by complete random assignment of groups or subjectsGroups are independentUsually employs strong controlQuasi-Experime

3、ntal DesignsGroups or subjects not randomly assignede.g., sample of convenienceMay not have a comparison groupTypical of clinical researche.g., within subjects repeated measuresLess “subject-intensive”Broad DistinctionsBetween subjectsDependent measures taken one timeData are independent (i.e., not

4、correlated)Within subjectsA “repeated measures” designDependent measures taken multiple timesData are dependentMixedBetween and withinDesign TypesSingle factor (one-way)Studies one independent variableMulti-factorStudies multiple independent variablesMay have several levelsExamples:Two-way (e.g., 2

5、x 2)Three-way (e.g., 2 x 2 x 2)Time-seriesSingle Factor DesignsPretest-posttest (one-group)Pretest-posttest (control group)Posttest-only (control group)Pretest-Posttest (one group)Quasi-experimentalOne set of measures taken before and after treatment or interventionCompare pretest and posttest score

6、sAnalysispaired t testWeaknessNo comparison or control groupPretest-Posttest (control group)Experimental design - random assignmentTwo groupsControlExperimentalMeasures on dependent variable made on both groups pre- and posttestSignificant differences in experimental group not found in control group

7、 attributable to treatmentAnalysisdifference scores compared with independent t testANCOVA pretest score as covariateMultiple Factor DesignsTwo-way factorial e.g., 2 x 3Three-way factoriale.g., 2 x 2 x 3Two-Way Factorial DesignStudies multiple independent variablesMain effects (ME)Each with a number

8、 of levels (L)Permits study of interactionsAnalysisANOVAExample: 2 x 3ME2L1 L2 L3L1L2ME1Three-Way Factorial DesignStudies multiple independent variablesMain effects (ME)Multiple levels (L)Interactions effectsAnalysisANOVAPost hoc pairwise comparisonsExample 2 x 2 x 3L1 L2 L3L1L2L2L1ME1ME 2ME3Counter

9、balanced DesignPossibility of order effects biasing data in a repeated measures designSolutionsRandomize orderCounterbalance trials - order systematically variedExample - two treatments (T1 - T2) “Crossover design”Half of subjects - T1 then T2Half of subjects - T2 then T1Latin Square DesignMinimizes

10、 order effectsA B CB C AC A B1 2 3Test sessionSubject 1Subject 2Subject 3Single Subject DesignPermits analysis of effects of treatment in individual subjects (or groups)ElementsSubjects usually own controlRepeated measuresDesign phases (times series analysis)Single Subject DesignTime series analysisDependent measure is continuousEstablish baselineMeasure treatment effect over timeBaselineTreatmentTimeCase ReportSubject a single individualOften uses a narrative formatMay be non-experimental or experimentalD