1、Lecture 14Understanding quantitative data Topics to be coveredPart 1Purpose of analysisPlanning the analysisUnderstanding dataTypes of data analysisPart 2Univariate descriptive analysisPart 3Bivariate descriptive analysisData reductionKey resourcesChapter 14, McGivern, 4eWebsites Overview of analysi

2、s processGo back to the brief and the proposalRe-familiarise yourself with what client needs to knowReview the data and select only data relevant to what client needs Work through the dataHighlight bits of interest/relevanceThink about organising/reducing mass of dataBuild up the storyPart 1Purpose

3、of analysisPlanning the analysisUnderstanding dataTypes of data analysisPurpose of analysisTo turn data into information and knowledgeTo extract meaningful insights from dataTo produce valid and reliable findings that address the clients problemPlanning the analysisReview The brief: what the client

4、asked forThe sampling plan: who you askedThe proposal: the research objectivesThe questionnaire: what you asked Write outWhat you need the analysis to tell youHow you are going to approach the analysisWhy plan?You will have lots of data from a questionnaireEven a short questionnaire among a small sa

5、mpleEasy to get overwhelmed by dataEasy to lose sight of aim of analysisBefore you start you need to be clear about the link between the analysis and the research objectivesthe research objectives and client problemUnderstanding dataThink back to questionnaire designConcepts translated into question

6、sDuring fieldwork Responses to questions recorded on questionnaires During data processingQuestions = variables, responses = valuesEach completed questionnaire = caseEach case has a set of variables and valuesLevels of measurementAll responses appear as number codes in data analysis grid/softwareNum

7、ber codes Different levels of meaning or measurementNominalOrdinal IntervalRatioImportant later in determining what analysis/statistics to useKey stages in data processingData entry/inputChecking and verifyingCleaning and editingCoding open-ended questionsDesigning and running tables, etc.Checking t

8、ablesData entry or data inputFor pen and paper questionnaires (PAPI)Data entered or keyed in by handScanned in electronically(if questionnaire designed appropriately)Computer-aided data capture (CAPI, etc.)Data entered as it is being collectedNo separate data entry stageData entry/analysis softwareE

9、xamples Specialist software, e.g. SPSSSpreadsheet, e.g. ExcelBasics of bothA grid or table layoutEntered case number by variable/valueThat is, respondent by responseTranslation of response to variable/value = codingData cleaning and editingWhy?To make sure dataset free of errors and inconsistenciesE

10、xamplesMissing valuesOut of range valuesErrors due to misrouting of questionsMake sure dataset clean before starting analysisTypes of data analysisFour main typesUnivariateBivariateExplanatoryInferentialPart 2 Univariate descriptive analysisUnivariate analysisAnalysis of one variableIncludes the fol

11、lowing:Frequency counts/frequency distributionProportions and percentagesMeasures of central tendencyMeasures of dispersion or variationWhy use univariate?Allows you to Describe the distribution and variation of the values of a variableSummarise the data from the whole sampleSee the bigger picture a

12、s well as the detailPrepare the data for further investigationGraphical displaysPie chartsGood for showing parts of a whole, e.g. number or % in each social classBar chartsUsed for nominal/ordinal level variables, e.g. number or % in each age bandHistogramsUsed for interval/ratio level variables, e.

13、g. number or % across age range (actual age)Line graphsFigure 14.3 Example of a pie chartFigure 14.4 Example of a simple bar chartFigure 14.8 Example of a histogramSome better examplesHave a look at these websitesGapminder Google visualisation IBMs Many Eyes Measures of central tendencyMeanArithmeti

14、c mean or averageUse for interval/ratio level variablesModeMost common valueUse for any level of variableMedianMiddle value when arranged in orderUse for all levels except nominalMeasures of variationRange Spread of values from highest to lowestUse for interval/ratio level variablesStandard deviatio

15、nSummarises spread around the meanUse for interval/ratio level variablesPart 3Bivariate descriptive analysisData reductionBivariate descriptive analysisAnalysis of two variablesRelationship or associationE.g. between gender and number of texts sentDivide the gender variable into twoMen vs. WomenDivi

16、de the number of texts sent per month into bands, e.g. less than 50; 5099; 99+TerminologyIdeas and hypothesesCross-tabulationsCross breaks, top headings and bannerBase size, filtering and weighting Dependent and independent variableIdeas and hypothesesWhen starting the analysisClients information ne

17、edsResearch objectivesQuestions from questionnaireIdeas, notions, etc., about relationships, etc.HypothesesThings that you want to check out in the dataSome you will have before you start and some will occur to you during analysisSee McGivern Case study 14.1Cross-tabulationsMost common way of doing

18、bivariate analysisTable row and column variablesColumn variables = cross break, top heading = independent or explanatoryRow variables = stub = dependent or e variableBox 14.4 Example of a cross-tabulation using demographic variablesBases and filtering Base sizeEach column based on total number in th

19、at group/those answering the questionIncluding Dont know responses in column percents?Filtering Using something other than total sample on which to base the columnExample: filter for responses from menWeightingWeighting used to adjust profile of sample dataTo make it representative on particular cha

20、racteristics of a target groupE.g. demographics and product usageSee McGivern Case study 14.2Cross-tabs can display weighted and unweighted dataDependent and independent variablesRelationships between variablesOne variable designated independentAlso known as explanatory variableOther variable design

21、ated dependent Also known as e variableBased on assumptions and prior knowledgePredict that the dependent variable will change as a result of the independentPatterns and relationshipsAim of analysis is to look for and examine patterns and relationships in the dataMain method Bivariate analysisPositi

22、ve, negative or no relationship?Use measures of associationWhich one depends on level of measurement of variables involvedMeasures of associationCramers VContingency coefficientPhiKendalls tau-bGammaSpearmans rhoPearsons rLimitationsBivariate descriptive analysisCan say something about relationshipsAn association between variablesDo say something about influence and direction of influenceDesignating one variable independent, other as dependentCannot say anything ab