講義講稿期末

1、PowerPoint to panyNaresh MalhotraJohn HallMike ShawPeter OppenheimPowerPoint to panyChapter 10Sampling Design and ProcedureChapter ObjectivesAfter reading this chapter, you should be able to:Differentiate a sample from a census and identify the conditions that favour the use of a sample versus a cen

2、sus Discuss the sampling design process: defining the target population, determining the sampling frame, selecting sampling technique(s), determining sample size and executing the sample processClassify sampling techniques as non-probability and probability techniques Describe the non-probability sa

3、mpling techniques of convenience, judgmental, quota and snowball samplingChapter Objectives (continued)Describe the probability sampling techniques of simple random sampling, systematic, stratified and cluster samplingIdentify the conditions that favour the use of non-probability sampling sampling v

4、ersus probability sampling Define key concepts and quantitative symbols pertinent to samplingDiscuss statistical to determining sample size based on simple random samplingUnderstand the other probability approaches to determining sample sizeExplain the use of computers in sampling designChapter Obje

5、ctives (continued)Understand the sampling design process and the use of sampling techniques in international marketing researchConsider the careers available in the market research industryDefine the seven steps of the marketing research processList the key outputs in the research processDescribe th

6、e activities involved in defining the management decision problemExplain why the management decision problem is the base on which marketing success is builtChapter OutlineSample or censusSampling design processA classification of sampling techniquesNon-probability sampling techniquesProbability samp

7、ling techniquesNon-probability versus probability samplingStatistical considerations in the sampling processDefinitionsSymbolsChapter Outline (continued)The sampling distribution of the meanStatistical approaches to determining the sample sizeThe confidence interval approachThe hypothesis testing ap

8、proachOther probability sampling techniquesAdjusting the statistically determined sample sizeInternational marketing researchComputer applicationsSummary TopicSample or censusSampling design processA classification of sampling techniquesNon-probability sampling techniquesProbability sampling techniq

9、uesNon-probability versus probability samplingStatistical considerations in the sampling processDefinitionsSymbolsSample or Census Sampling is one of the elements of research design involving several basic questionsShould a sample be taken?If so, what process should be followed?What kind of sample s

10、hould be taken?How large should it be?What can be done to control and adjust for non-response errors?Sample or Census (continued)The objectives of most marketing research projects is to obtain information about the characteristics of a population, eg:Proportion of loyal customersSuch information can

11、 come from a census or a sampleCensus complete enumeration of the elements of a populationSample subgroup of the population selected for participation in a studySample characteristics are called “statistics” which we use to make inferences about the population estimation procedures & hypothesesCondi

12、tions Favouring the Use of Sample vs CensusSampleCensusBudgetTime availablePopulation sizeVariance in the characteristicsCost of sampling errorsCost of non-sampling errorsNature of measurementAttention to individual casesSmallShortLargeSmallLowHighDestructiveYesLargeLongSmallLargeHighLowNon-destruct

13、iveNoTopicSample or censusSampling design processA classification of sampling techniquesNon-probability sampling techniquesProbability sampling techniquesNon-probability versus probability samplingStatistical considerations in the sampling processDefinitionsSymbolsThe Sampling Design ProcessIt is th

14、e collection of elements or objects that possesses the information sought by the researcher and about which inferences are to be made Precise statement of who should and should not be included in the sampleElement:An object about which, or from which the information is desired usually the respondent

15、Define the Target PopulationDefine the Target Population (continued)Sampling unit:An element or a unit containing the element that is available for selection at some stage of the sampling processExample Revlon wants to sample girls over 18 years old:Sample girls directly sampling unit = the elementS

16、ample households (sampling unit) and then interview all 18 year old girls (element) from the householdDefine the Target Population (continued) Extent:Geographical boundaries:eg western metropolitan region of Melbourne, national study, study of two countries (Malaysia and Australia)Time:Period under

17、considerationDefining the Target PopulationThis can be difficult, eg assessing consumer response to a new brand of mens aftershaveWho should be included?All men?Men who have used an aftershave in the past month?Men 17 and older?Women as they buy it as a present?Determine the Sampling FrameA sampling

18、 frame is a representation of the elements of the target population. It is a list or set of directions for identifying the target populationTelephone book white or yellow pagesAn association directory MRSA list of research organisations or membersMailing list purchased from a commercial business, me

19、mbership listCity directory or map Random digit dialling RDDSampling Frame ErrorA list that may omit some elements of the population or include other elements which do not belong leads to sampling frame errorPerth WhitepagesTM may omit people with an incorrect listing, silent numbers, and people out

20、side the metropolitan areaIf our target population is people who purchased car tyres in the last 3 months, the WhitepagesTM would also include people who have not purchased car tyres (in the last 3 months)Discrepancy may be so small it can be ignored or managedScreen respondents or adjust data by a

21、weighting schemeSelect a Sampling TechniqueSeveral broad decisions:Bayesian v traditionalWith or without replacementBayesianElements are selected sequentiallyAfter each element is added to the sample, data is collected, sample statistics computed, sampling costs determined Not used widely in marketi

22、ng research costs & probabilities not availableTraditionalEntire sample is selected before data collection begins focus of following sectionsSelect a Sampling Technique (continued)Sampling with replacementAn element is selected from the sampling frame and appropriate data obtainedElement is placed b

23、ack in the sampling frame:Hence can be included more than onceSampling without replacementOnce an element is selected for inclusion in the sample, it is removed from the sampling frame and therefore cannot be selected againProbability v Non-probabilityMost important sampling decisionDetermine the Sa

24、mple SizeDetermine the number of elements to be included in the studyQualitative factors:Importance of the decision more important than more precision needed hence larger samplesNature of the researchNumber of variables use small number for exploratory research, larger number for conclusive research

25、 and where there are numerous variablesNature of the analysis larger sample where multivariate statistics, great detail needed or more granular analysis undertakenDetermine the Sample SizeQualitative factors:Sample size used in similar studiesResource constraints money, people & timeIncidence rates

26、how many need to be contacted to obtain the required sample sizeCompletion ratesAnticipated refusalsIncidence rate:Percentage of people eligible to participateDetermines number of contacts need to be screened for a given sample size requiredExample 10.2 on p. 367 illustrates this processExecute the

27、Sampling ProcessRequires detailed specification of how the sampling design decisions:PopulationSampling frameSampling unitSampling techniqueSample sizeare to be implementedIf households are the sampling unit then definition of household needed:Procedures needed for vacant houses and callbacks for th

28、ose not at homeTopicSample or censusSampling design processA classification of sampling techniquesNon-probability sampling techniquesProbability sampling techniquesNon-probability versus probability samplingStatistical considerations in the sampling processDefinitionsSymbolsFigure 10.2: Classificati

29、on of Sampling TechniquesSampling TechniquesNon-probabilityPersonal judgement of the researcher is used rather than chance to select elementsDifficult to generalise result to the populationUsed in studies where projection to the population is not necessary:Concept tests, package tests, and copy test

30、s Sampling Techniques (continued)ProbabilitySampling units are selected by chancePre-specifying every potential sample of a given size that could be drawn from the populationRequire precise definition of the target population and sampling frameAble to make inferences about the target populationUsed

31、when there is a need to estimate market share or provide information on product category, brand usage rates, psychographic and demographic profiles of usersTopicSample or censusSampling design processA classification of sampling techniquesNon-probability sampling techniquesProbability sampling techn

32、iquesNon-probability versus probability samplingStatistical considerations in the sampling processDefinitionsSymbolsConvenience SamplingSelection of sampling units is left to the interviewer right place, right time accessibilityInexpensive, quick can be used for exploratory research Selection bias p

33、resent (self-selection) not representative, cannot generalise to the populationAppropriate for exploratory research generate hypotheses, theories, insights, etc:Students at uni, shopping centres without qualifying respondents, questionnaires in magazines or restaurantsExample 10.4 p. 369 illustrates

34、 convenience samplingJudgmental Sampling (Form of Convenience)Selection based on researcher judgement of whether the elements believe they are, eg representative of the populationInexpensive, convenient, quick Cannot generalise to specific populationsSelection of test markets, expert court witnesses

35、, dept stores selected to test new merchandising displaysQuota SamplingTwo-stage restricted judgmental sampling1st develop quotas based on relevant characteristics and determine the distribution in relation to the population proportion eg Age: 18 25 (25%),2nd sample elements are then selected based

36、on convenience or researcher judgementMay not be representative of the population but could be relevantSelection and self-selection bias possibleLower cost and greater convenienceExample 10.5 on p. 370 illustrates quota samplingSnowballing SamplingInitial group of respondents is selected at random,

37、then asked to identify others who belong to the target population of interestReferrals will have demographic and psychographic characteristics that are more similar to person referring than would be by chance:Minority groups, widowed males under 35, people involved in a specialised craft Substantial

38、 increased likelihood of locating desired sample, results in low sampling variance and cost Example 10.6 on p. 370 illustrates snowball samplingTopicSample or censusSampling design processA classification of sampling techniquesNon-probability sampling techniquesProbability sampling techniquesNon-pro

39、bability versus probability samplingStatistical considerations in the sampling processDefinitionsSymbolsProbability Sampling TechniquesThese vary in terms of sampling efficiencyWhich is a trade-off between sampling cost and precisionPrecision is the level of uncertainty & is inversely related to sam

40、pling errors but directly related to costResearcher should strive for the most efficient sampling design given budget restraintsEfficiency of any probability sampling technique should be compared to simple random sampling (SRS)Simple Random SamplingEach element in the population has a known and equa

41、l chance of selectionA sample is drawn by a random procedure from a sampling frameEasily understood just like a lotteryGeneralisation to the population is possibleDifficult to construct a sampling frameSamples may be spread over large geographical areas, hence high time and cost in data collectionLo

42、wer precision (large standard errors) May or may not result in representative sampleNot widely used in MRStratified Sampling (continued)Two-step processPopulation split into sub-populationsStrata are mutually exclusive and collectively exhaustiveThen SRS from each stratum to select the elements:With

43、in stratum homogeneousBetween strata heterogeneousVariables used for stratification:DemographicCustomer typeFirm size or industryAge: 18 - 25 year olds would have similar characteristics whereas 46 - 55 year olds would have another setStratified SamplingProportionate vs disproportionate?Proportionat

44、e size of sample drawn from each stratum is proportionate to the relative size of that stratum to the total population & to the standard deviation of the distribution of the characteristic of interestDisproportionate larger strata will be relatively more important in determining population mean, etc

45、 hence so it should be with sampling. To increase precision more elements drawn from strata with larger standard deviations. Knowledge of distribution of characteristic of interest needed not always the caseCluster SamplingTarget population is split into mutually exclusive and collectively exhaustiv

46、e sub-populations (sampling frame)Then random sample of clusters is selected based on SRS All elements or a sample from each (selected) cluster is then selected:Within cluster - homogeneousBetween clusters - heterogeneousCommon form is area sampling Figure 10.3 Types of Cluster SamplingDifferences B

47、etween Stratified and Cluster SamplingOnly one sample of subpopulations (cluster) is chosen, whereas all subpopulations (strata) are selected for further sampling The objective of cluster sampling is to increase efficiency by decreasing costs, whereas the objective of stratified sampling is to incre

48、ase precision Homogeneity and heterogeneity criteria are differentExample 10.7 p. 376 illustrates two-stage area samplingOther Probability Sampling TechniquesSequential samplingPopulation elements sampled sequentially, data collection & analysis done at each stage & then decision to sample any more

49、madeTesting preference for an item continue to sample until confident that preference establishedDouble sampling (aka two-phase)Certain population elements sampled twiceSome info from all elements in 1st sampleExtra info from sub-sample in 2nd phase (when more is known about the sample)Refer to Figu

50、re 10.4 p. 378 for a description of procedures for drawing probability samplesStrengths and Weaknesses of Basic Sampling TechniquesTable 10.2TopicSample or censusSampling design processA classification of sampling techniquesNon-probability sampling techniquesProbability sampling techniquesNon-probab

51、ility versus probability samplingStatistical considerations in the sampling processDefinitionsSymbolsNon-probability Versus Probability SamplingUses of Non-probability and Probability SamplingNon-probability sampling used inConcept testsPackage testsName testsCopy testsProbability sampling used when

52、 high need for accuracyMarket shareSales volumesGeneralisation to population not criticalCentral location intercept quota samplingTelephone survey (RDD) with either stratified or systematic samplingTopicSample or censusSampling design processA classification of sampling techniquesNon-probability sam

53、pling techniquesProbability sampling techniquesNon-probability versus probability samplingStatistical considerations in the sampling processDefinitionsSymbolsDefinitionsDefinitions (continued)TopicSample or censusSampling design processA classification of sampling techniquesNon-probability sampling

54、techniquesProbability sampling techniquesNon-probability versus probability samplingStatistical considerations in the sampling processDefinitionsSymbolsSymbolsTopicThe sampling distribution of the meanStatistical approaches to determining the sample sizeThe confidence interval approachThe hypothesis

55、 testing approachOther probability sampling techniquesAdjusting the statistically determined sample sizeInternational marketing researchComputer applicationsSummary The Sampling Distribution of the MeanDistribution of the values of a sample statistic computed for each possible sample that could be d

56、rawn from the target population under a specified sampling planStatistical inference:The process of generalising the sample results to the population resultsThe Sampling Distribution of the Mean (continued)Important properties of the sampling distribution of the mean (and proportion) for large sampl

57、es (30+):Normal distributionMean of sampling distribution of the mean equals the corresponding population parameter valueStandard deviation (SD) is called the standard error (SE)When population SD not known it can be estimatedSE of the proportion can also be estimatedThe Sampling Distribution of the

58、 Mean (continued)Area under the sampling distribution between any two points can be calculated in the terms of z values:Number of SEs a point is away from the meanWhen sample size is 10%+ of the population size, the SE formulae will overestimate the SD of the population mean or proportionRefer to pp

59、. 383-4 for a more complete description of these properties and for the relevant statistical formulaeTopicThe sampling distribution of the meanStatistical approaches to determining the sample sizeThe confidence interval approachThe hypothesis testing approachOther probability sampling techniquesAdju

60、sting the statistically determined sample sizeInternational marketing researchComputer applicationsSummary Statistical Approaches to Determining the Sample SizeBased on traditional statistical inference the precision level is specified in advanceTopicThe sampling distribution of the meanStatistical