Measurement And Scaling - Åbo Akademi

Transcription

Measurement and scalingFöretagsakademin, Henriksgatan 7 FIN-20500ÅboMeasurement and scaling Measurement: assigningnumbers or othersymbols tocharacteristics ofobjects according tocertain pre-specifiedrules Scaling: generating acontinuum upon whichmeasured objects arelocated Primary scales (p.237)– Nominal scale numbers assigned tofootball players social securitynumbers brand numbers, storytypes, sexclassifications percentages mode Chi-square,binominal testsFöretagsakademin, Henriksgatan 7 FIN-20500 ÅboMeasurement and scales Ordinal scale Interval scale– rank order or winner– quality ranking– preference rankings,market position, socialclass– percentile, median– rank-order, correlation,Friedman ANOVA– performance rating on a0-10 scale; 8.2; 9.2; 9.6;temperature– attitudes, opinions– range, mean, std– product-moment,correlations, t-tests,ANOVA, regression,factor analysisFöretagsakademin, Henriksgatan 7 FIN-20500 Åbo1

Measurement and scales Ratio scale time to finish, in seconds. Zero point is fixedlength, weightAge, income, costs, sales, market shareCoefficient of variationFöretagsakademin, Henriksgatan 7 FIN-20500 ÅboScaling techniques Comparative scales– direct comparison ofstimulus objects– do you prefer Cokeor Pepsi– ordinal or rankorder properties– no-metric scaling paired comparison,rank order, constantsum scales, Q-sort Non-comparative scales– each stimulus object isscaled independently ofothers– interval or ratio scales– continuous ratings oritemized rating scales Likert, semantic Most widely usedscaling technique usedin marketing researchFöretagsakademin, Henriksgatan 7 FIN-20500 ÅboNon-comparative scaling techniques Continuous Rating Scales– on a scale between 0 (very low) and 100(very high) indicate what is the probabilitythat you will buy a car next year Itemized Rating Scales– a measurement scale having numbersand/or brief descriptions associated witheach category. The categories are orderedin terms of scale positionFöretagsakademin, Henriksgatan 7 FIN-20500 Åbo2

Itemized Rating Scales Likert scale– 5 point scale from 1-5strongly disagree tostrongly agree– can be 7 point scale Semantic Differential– a 7-point rating scalewith end pointsassociated with bipolarlabels that have semanticmeaningCold--------WarmPowerful-------Weak Semantic differentials arepopular in marketing– comparing brand, product,and company images– for developing advertisingand promotion strategies,– new product developmentstudies Stapel scale– measure attitudes, evennumber range without zerofrom -5 to 5Företagsakademin, Henriksgatan 7 FIN-20500 ÅboNon-comparative itemized rating scaledecisions The number of scale categories to use balanced versus unbalanced scales Odd or even number categories forced versus non-forced choices the nature and degree of the verbaldescription the physical form of the scaleFöretagsakademin, Henriksgatan 7 FIN-20500 ÅboQuestionnaire and Form DesignFöretagsakademin, Henriksgatan 7 FIN-20500Åbo3

Questionnaire design process Specify informationneeded Specify the type ofinterview method Determine the contentof each Q Design the Qs toovercome therespondent’s inability anunwillingness to answer Design on the Qstructure Determine the wording Arrange the Qs inproper order Identify the form andlayout Reproduce PretestFöretagsakademin, Henriksgatan 7 FIN-20500 ÅboIs the Q necessary or do you need morethan one Q Do you think Cokeis tasty andrefreshing Avoid why questions!– They tend to containtwo aspects: attributes of theproduct influences leading toknowledge thereof– Wrong! DoubleBarreled Q Do you think C istasty? Do you think C isrefreshing?– Why do you shop atStockmann - Wrong!– Instead: What do youlike about Stockmann ascompared to otherstores?– How did you first happento shop at S?Företagsakademin, Henriksgatan 7 FIN-20500 ÅboDon’t make respondents remember People have a poormemory– What is the brand nameof the shirt you wore 2weeks ago?– What did you have forlunch a week ago?– What were you doing amonth ago at noon?– How many liters of softdrink have youconsumed during the lastmonth? Ask instead How often do you drinksoft drinks in a typicalweek––––less tahn once a week1 to 3 times per week4 to 6 times per week7 or more times a week Telescoping(compressing time) Creation (did notoccur)Företagsakademin, Henriksgatan 7 FIN-20500 Åbo4

People don’t want to answer sensitiveinfo accurately Put sensitive topics at moneythe end of the family lifeQuestionnaire political and religious Ask the Q in thirdbeliefsperson Hide the Q in a group involvements inof Qs that theaccidents or crimesrespondent is willing toanswer Provide responsecategoriesFöretagsakademin, Henriksgatan 7 FIN-20500 ÅboOrder or position bias The respondent’s tendency to check analternative based on its position in a listof alternatives.– Respondents tend to check the first or thelast - especially the first– Sometimes also the middleFöretagsakademin, Henriksgatan 7 FIN-20500 ÅboChoose Q wording Use ordinary words Make sure the word has a singlemeaning. Problematic ones are ‘usually’,‘normally’, ‘frequently’, ‘often’,‘regularly’, ‘occasionally’, and‘sometimes’ Avoid leading or biased Qs Questionnaire Design Checklist.Malhotra pp. 304-305!Företagsakademin, Henriksgatan 7 FIN-20500 Åbo5

Sample Population - the aggregate of all theelements that share some common setof characteristics– the proportion of consumers who are loyalto a particular brand of toothpaste Census - a complete enumeration of theelements of a population Sample - a subgroup of the populationselected for participation in the studyFöretagsakademin, Henriksgatan 7 FIN-20500 ÅboThis you need to know by heartMARKETThe set of all actual and potential buyers of aproductPOTENTIAL The set of consumers who profess a sufficientMARKET level of interest in the market offerAVAILABLE The set of consumers who have interest, income,MARKET and access to a particular market offerTARGETMARKETMARKETDEMANDThe part of the available market the companydecides to pursueThe total volume of a product that would bepurchased by a defined customer group in adefined geographical area in a defined timeperiod in a defined marketing environment undera defined marketing programFöretagsakademin, Henriksgatan 7 FIN-20500 ÅboFöretagsakademin, Henriksgatan 7 FIN-20500 Åbo6

Now a few moreSampleA subgroup of the elements of the populationselected for participation in the studyTargetPopulationThe collection of elements or objects thatpossess the information sought by the researcherElementObjects that possess the information sought forby the researcherSamplingUnitThe basic unit containing the elements of thepopulation to be sampledSamplingframeA list or set of directions for identifying the targetpopulation, e.g. telephone book, listing of firms inan industry, a city directory.Företagsakademin, Henriksgatan 7 FIN-20500 ÅboSample sizes used in marketing researchType of studyMinimumTypical RangeProblem identification(market potential)5001000-2500Problem solving(pricing)200300-500Product test200300-500Test-marketing 00Test-market audits10 stores10-20 storesFocus groups6 groups10-15 groupsFöretagsakademin, Henriksgatan 7 FIN-20500 ÅboAnswers to a single variable How many users of abrand can beconsidered brand loyal? What percentage of themarket can beconsidered as heavyusers, medium users,light users, andnonusers? How many customers arefamiliar with a new product;somewhat familiar, unfamiliarwith the brand? What is themean familiarity rating? Isthere much variance in theextent to which customersare familiar with the newbrand? What is the incomedistribution of brand users?Is this distribution skewedtoward low-incomebrackets?Företagsakademin, Henriksgatan 7 FIN-20500 Åbo7

Frequency distribution One variable isconsidered at a time Measures of locationtend to measure wherethe center is: Measure of shape:– skewness (the symmetryof the mean)– -0.094 would indicate aslight negative skew– Kurtois(the relativepeakedness or flatness ofthe curve; -1.261indicates the distributionbeing a bit flatter thannormal– Mean– Mode (the value thatoccurs most frequently)– Median– Range– Variance & STDFöretagsakademin, Henriksgatan 7 FIN-20500 ÅboHypothesis testing Formulate the nullhypothesis– a statement whereno difference oreffect is expected Formulate thealternativehypothesis– statement wheresome difference isexpected Type I error: whenthe results lead tothe rejection of thenull hypothesis whenit is in fact true. The probability oftype I error ( ) iscalled level ofsignificanceFöretagsakademin, Henriksgatan 7 FIN-20500 ÅboHypothesis testing Type II errors occur If is extremely lowwhen the null(0.001) type IIhypothesis is noterrors are veryrejected when it is inlikely. Normally, isfact falseset at 0.05 or 0.01 The power of a testis the probability ofrejecting the nullhypothesis when itshould be rejectedFöretagsakademin, Henriksgatan 7 FIN-20500 Åbo8

Cross-tabulation Answering questions tohow the single variableis related to othervariables. How many bran-loyalusers are male? Is product use (heavyusers, medium users,light users, nonusers)related to interest inoutdoor activities (high,medium, low)? Is familiarity with anew product relatedto age andeducation? Is productownership related toincome (high,medium, low)?Företagsakademin, Henriksgatan 7 FIN-20500 ÅboCross-tabulation also called contingency tables cross-tabulation with two variables isalso called bivariate cross-tabulation The general rule is to compute thepercentages in the direction of the ivacross the dvFöretagsakademin, Henriksgatan 7 FIN-20500 ÅboCross-tabulationInternet usageGenderInternet 100%100% The dv is internet usage; the iv is gender. Assuming the otherway makes no sense, i.e. that heavy internet users causes peopleto be male!! But maybe this finding can be mediated by a third variable? Age?Företagsakademin, Henriksgatan 7 FIN-20500 Åbo9

Cross-tabulation - a third variablePurchase of fashion clothingmaritial otal100%100% Refine the association between the original twocan indicate no association between the 2 variables, although initiallyobservedcan reveal some association between the two variables although notinitially observed - a suppressed association between the 2 v:sIndicate no change in the initial associationFöretagsakademin, Henriksgatan 7 FIN-20500 ÅboCross-tabulation - a third variableGenderPurchase offashionclothesMale marital statusFemale marital %60%Low65%60%75%40%Total100%100%100%100% 60% of the high purchasers are unmarried female, 25% of thosewho are married Differences among the male far less Gender has refined the original result Unmarried respondents are more likely to be high-purchasersand this effect is more pronounced for females than malesFöretagsakademin, Henriksgatan 7 FIN-20500 ÅboCross-tabulation The statisticalsignificance of theobserved association iscommonly measyredwith chi-square The strength of theassociation is interestingonly if it is significantand can be measuredwith phi correlationcoefficient, Cramer’s Vor lambdaFöretagsakademin, Henriksgatan 7 FIN-20500 Åbo10

Measurement and scaling Företagsakademin, Henriksgatan 7 FIN-20500 Åbo Measurement and scaling Measurement: assigning numbers or other symbols to characteristics of objects according to certain pre-specified rules Scaling: generating a continuum upon which measured objects are located Primary scales (p. 237) –Nominal scale numbers assigned to