Sunteți pe pagina 1din 15

CHAPTER 10 MEASUREMENT AND SCALING: NONCOMPARATIVE SCALING TECHNIQUES OPENING QUESTIONS 1. 2. ". #. %. '. (. ). *.

How are noncomparative scaling techniques different from comparative scaling, and what is the distinction between continuous and itemized rating scales? What are the differences between Li ert, semantic differential, and !tapel scales? What are the decisions involved in constructing itemized rating scales, and what options should be considered? How are scales evaluated, and what is the relationship between reliabilit$ and validit$? &n what wa$ can noncomparative scaling contribute to total qualit$ management? What considerations are important in implementing noncomparative scales in an international setting? How does technolog$ affect noncomparative scaling? What ethical issues are involved in developing noncomparative scales? How can the &nternet be used to construct noncomparative scales?

AUTHOR'S NOTES: CHAPTER FOCUS +his chapter provides a discussion of the noncomparative scales. ,ontinuous and itemized rating scales are discussed. +he important noncomparative itemized rating scale decisions are e-amined and guidelines provided. +he construction of multi.item scales is described. +he evaluation of scales in terms of measurement accurac$, reliabilit$, validit$, and generalizabilit$ is discussed at some length. +he discussion of scaling techniques is much more e-tensive as compared to competing te-ts. /-amples, including instructions for administering these scales, have been provided to illustrate the various scales. +his chapter could be taught b$ focusing on the opening questions sequentiall$. 0reater emphasis could be placed on continuous and itemized rating scales 12pening 3uestion 14, Li ert, semantic differential, and !tapel scales 12pening 3uestion 24, noncomparative itemized rating scale decisions 12pening 3uestion "4 and scale evaluation 12pening 3uestion #4. /thical issues in developing noncomparative scales 12pening 3uestion )4 also deserve special attention. OUTLINE 1. 2. ". 2verview 5oncomparative !caling +echniques i. ,ontinuous 6ating !cale &temized 6ating !cales i. Li ert !cale ii. !emantic 7ifferential !cale iii. !tapel !cale 5oncomparative &temized 6ating !cale 7ecisions i. 5umber of !cale ,ategories ii. 8alanced 9ersus :nbalanced !cales 12)

#.

%. '.

(. ). *. 1?. 11. 12. 1". 1#.

iii. 2dd or /ven 5umber of ,ategories iv. ;orced or 5onforced ,hoice v. 5ature and 7egree of 9erbal 7escription vi. <h$sical ;orm or ,onfiguration =ulti.item !cales !cale /valuation i. 6eliabilit$ ii. +est.6etest 6eliabilit$ iii. >lternative.;orms 6eliabilit$ iv. &nternal ,onsistenc$ 6eliabilit$ v. 9alidit$ a. ,ontent 9alidit$ b. ,riterion 9alidit$ c. ,onstruct 9alidit$ vi. 6elationship 8etween 6eliabilit$ and 9alidit$ ,hoosing a !caling +echnique !ummar$ &llustration :sing the 2pening 9ignette &nternational =ar eting 6esearch +echnolog$ and =ar eting 6esearch /thics in =ar eting 6esearch !ummar$ @e$ +erms and ,oncepts >cron$ms

TEACHING SUGGESTIONS 2pening 3uestion 1 A 7escribe the different noncomparative scaling techniques. &f available, bring e-amples of the different scales to class to show to the students. 8egin b$ defining noncomparative scaling as the t$pe of scaling which does not compare the obBect against another obBect or some standard. 6ather, the rater uses whatever standard seems most appropriate to him or her. 1. ,ontinuous rating scaleC the respondents rate the obBects b$ placing a mar at the appropriate position on a line that runs from one e-treme of the criterion variable to the other. +he form of the continuous scale varies considerabl$ depending on the imagination of the researcher. +heir use in mar eting has been limited because the$ are not as reliable as itemized scales, the scoring process is cumbersome, and the$ provide little additional information. 2. &temized rating scaleC the respondents are provided with scales having numbers andDor brief descriptions associated with each categor$. +he respondents are required to select one of the specified categories that best describes the obBect being rated. Opening Ques i!n " A 7iscuss the various t$pes of itemized rating scales. 12*

Li ert scaleC the respondents are required to indicate a degree of agreement or disagreement with each of a series of statements related to the stimulus obBects. +he Li ert scale is often used in mar eting. &t is eas$ to construct and administer, it is eas$ for respondents to complete, and it is suitable for mail, telephone, and personal surve$s. !emantic differential scaleC is a seven.point rating scale with end points associated with bipolar labels that have semantic meaning. 6espondents are required to rate obBects on a number of itemized, seven.point rating scales bounded at each end b$ one of two bipolar adBectives. +his scale is popular in mar eting and has been used in image studies, promotion strateg$, and new product development studies. !taple scaleC is a unipolar rating scale with ten categories numbered from .% to E%, without a neutral point. 6espondents are as ed to indicate how accuratel$ or inaccuratel$ each term describes the obBect b$ selecting an appropriate numerical response categor$. +hough easier to construct than the !emantic 7ifferential, while giving the same results, the !taple scale has not been widel$ applied in mar eting. 2pening 3uestion " A Highlight the maBor decisions involved in constructing itemized rating scales. :se a running e-ample li e the 7epartment !tore <atronage <roBect to aid in student understanding. 1. 5umber of scale categories to useC the number of categories should be between five and nineF however, there is no single, optimal number of categories, which would be applicable for all scaling situations. 2. 8alanced versus unbalanced scaleC a balanced scale has an equal number of favorable and unfavorable categories usedF otherwise, the scale is unbalanced. +he scale should be balanced in order to obtain obBective dataF however, if the distribution of responses is li el$ to be s ewed, an unbalanced scale with more categories in the direction of s ewness ma$ be appropriate. :se ;igure 1?.#. Have the students discuss an$ potential s ewness in the responses. ". 2dd or even number of categoriesC with an odd number of categories, the middle scale position is generall$ designated as neutral. &f a neutral or indifferent scale response is a possibilit$ for at least some of the respondents, an odd number of categories should be used. !tudents can debate the effect of a neutral categor$ on the data obtained. #. ;orced versus non.forced scalesC a forced rating scale does not have a Gno opinionG or Gno nowledgeG option. +hus, the respondents without an opinion ma$ mar the middle scale position. &f a sufficient proportion of the respondents in fact do not have an opinion on the topic, mar ing the middle position in this manner will distort measures of central tendenc$ and variance. &n situations where the respondents are e-pected to have no opinion, the accurac$ of the data ma$ be improved b$ having a nonforced rating scale, which includes a Gno opinionG or Gno nowledgeG categor$. !tudents should discuss the appropriateness of a forcedDnonforced rating scale. %. 5ature and degree of verbal descriptionC the strength of the adBectives used to anchor the scale has a slight tendenc$ to influence the distribution of the responses. With strong anchors, respondents are less li el$ to use the e-treme scale categories. Have the students reach consensus on the scale anchors. +r$ to encourage multiple anchors, which can be used in the surve$. '. <h$sical form of the scaleC there is no agreement as to which form is the most appropriate, but scales could be presented verticall$ or horizontall$, categories could be e-pressed in terms 1"?

of bo-es, discrete lines, or units on a continuum and ma$ or ma$ not have numbers assigned to them, and numerical values could be positive or negative or both. +he students should decide which format to use for the scales. Have them Bustif$ their reasons for the scale the$ choose. !ummarize the discussion on rating scale decisions using +able 1?.2. 2pening 3uestion # A /-plain the criteria used to evaluate a multi.item scale. 1. =easurement accurac$C refers to capturing the responses as the respondent intended them to be understood. /rrors can result from either s$stematic error, which affects the observed score in the same wa$ on ever$ measurement, or random error, which varies with ever$ measurement. 2. 6eliabilit$C refers to the e-tent to which a scale produces consistent results if repeated measurements are made on the characteristic. +hus, the scale is free from random error. ". 9alidit$C refers to the accurac$ of measurement. 9alidit$ of a scale ma$ be defined as the e-tent to which differences in observed scale scores reflect true differences among obBects on the characteristic being measured, rather than s$stematic or random errors. #. 0eneralizabilit$C refers to the e-tent to which one can generalize from the observations at hand to the set of all conditions of measurement over which the investigator wishes to generalize, called the universe of generalization. ;igure 1?.' can be used to discuss the issues involved in scale evaluation. A /-plain how validit$ is measured. +o address this topic, $ou must first distinguish the t$pes of validit$C content, criterion and construct. +his underscores the point that measurement and theor$ are ine-tricabl$ intertwined. 1. ,ontent 9alidit$ . > subBective evaluation b$ e-perts in the domain being studied to determine if all relevant items are covered in the stud$. +hus, in the auto industr$ for e-ample, auto e-perts 1e-perienced professionals, professors, etc.4 would evaluate the scale on the items it covers vis.H.vis what the$ consider relevant real world factors. 2. ,riterion 9alidit$ . > comparative stud$ of the chosen measurement scaleIs performance in relation to other variables selected as meaningful is conducted. +his can be either concurrent or predictive. +hus, two measurements are made of the same variable and the results compared for consistenc$. ". ,onstruct 9alidit$ . +his entails a theoretical investigation of the construct the scale is measuring. &t tries to assess wh$ the scale wor s and the nature of the theor$ underl$ing the scale. ,onvergent, discriminant, and nomological validit$ are assessed. +he e-ample on self. concept in the te-t will prove helpful.

2pening 3uestion % A 7iscuss the !/693:>L scale for measuring qualit$.

1"1

!/693:>L, a multi.item, noncomparative scale, was developed to measure service qualit$. +he original !/693:>L proposed 1? dimensions on which service qualit$ was ratedC reliabilit$, responsiveness, competence, access, courtes$, communication, credibilit$, securit$, understandingD nowing the customer, and tangibles. +hese 1? dimensions were later combined into fiveC tangibles, reliabilit$, responsiveness, assurance, and empath$. !/693:>L can be divided into two sections. +he first section collects data about respondentsJ feelings toward qualit$ firms possessing certain features or attributes. +he second section as respondents to rate to what e-tent the$ believe a firm has certain attributes. +he answers from the first section give a measure of a consumerIs e-pectations for certain attributes for a qualit$ firm. +he second section measures the consumersJ perceptions for a given firm on the same attributes. !/693:>L uses the difference between the perception and e-pectation measures as a measure of qualit$. 2pening 3uestion ' A 7iscuss pan.cultural scales vs. scales that use a self.defined cultural norm. +he pan.cultural approach is used to develop scales, which are free of cultural biases. 2f the scaling techniques we have considered, the semantic differential scale ma$ be said to be pan. cultural. &t has been tested in a number of countries and has consistentl$ produced similar results. >n alternative approach is to develop scales that use a self.defined cultural norm as a base referent. +his approach is useful for measuring attitudes that are defined relative to cultural norms 1e.g., attitude toward marital roles4. &n developing response formats, verbal rating scales appear to be the most suitable. A &dentif$ other international concerns. K K K K !pecial attention should be devoted to determining equivalent verbal descriptors in different languages and cultures. &t is important that the scale end points and the verbal descriptors be emplo$ed in a manner that is consistent with the culture. &n designing the scale or response format, respondentsI educational or literac$ levels should be ta en into account. &t is critical to establish the equivalence of scales and measures used to obtain data from different countries.

2pening 3uestion ( A 7iscuss the role of technolog$ in facilitating noncomparative scaling.

7atabase managers, such as d8>!/, allow researchers to develop scales and test their appropriateness for a particular application. !pecialized programs, such as >++&+:7/ !,>L/! b$ <ersimmon !oftware, construct a variet$ of rating scales for measuring attitudes in mar eting and opinion research. /LW6&+/6 b$ ,omputers for =ar eting ,orporation 1,f=,4 of !an ;rancisco can customize scales for printed questionnaires or for use b$ telephone interviewers at computer screens in a fraction of the time this would ta e without automation. >nother technological development is MsmartN instruments that can constantl$ monitor their own condition and the qualit$ of the information the$ provide. +he$ can also Mtal N directl$

1"2

to the other components of the measurement process, ma ing the integration and processing of information quic and reliable. 7iscuss the 2ption ;inder. 2pening 3uestion ) A 7iscuss the ethical implications of misusing scale descriptors . /thical issues can arise in the construction of noncomparative scales. ,onsider, for e-ample, the use of scale descriptors. +he descriptors used to frame a scale can be manipulated to bias results in an$ direction. > researcher who wants to proBect the clientIs brand favorabl$ can as respondents to indicate their opinion of the brand on several attributes using seven.point scales framed b$ the descriptors from e-tremel$ poor to good. :sing a strongl$ negative descriptor with onl$ a mildl$ positive one has an interesting effect. >s long as the product is not the worst, respondents will be reluctant to rate the product e-tremel$ poorl$. &n fact, respondents who believe the product to be onl$ mediocre will end up responding favorabl$. +r$ this $ourself. How would $ou rate 8=W automobiles on the following attributes? 6eliabilit$ <erformance 3ualit$C <restigeC Horrible 9er$ poor 2ne of the worst 9er$ low 1 1 1 1 2 2 2 2 " " " " # # # # % % % % ' ' ' ' ( ( ( ( 0ood 0ood 0ood 0ood

+hus, we see how important it is to use balanced scales with comparable positive and negative descriptors. When this guide is not practiced, responses are biased and should be interpreted accordingl$. A &dentif$ other areas of ethical concern. /thical concerns also underscore the need to adequatel$ establish the reliabilit$, validit$, and generalizabilit$ of scales before using them in a research proBect. !cales that are invalid, unreliable, or not generalizable to the target mar et provide the client with flawed results and misleading findings, thus raising serious ethical issues. 2pening 3uestion * A 7iscuss the use of the &nternet in scale development and testing. ,ontinuous rating scales ma$ be easil$ implemented on the &nternet. +he cursor can be moved on the screen in a continuous fashion to select the e-act position on the scale which best describes the respondentJs evaluation. =oreover, the scale values can be automaticall$ scored b$ the computer, thus increasing the speed and accurac$ of processing the data. !imilarl$, it is also eas$ to implement all three itemized rating scales on the &nternet. =oreover, using the &nternet, one can search for and locate similar scales used b$ other researchers. +he 2ffice of !cales 6esearch at !outhern &llinois :niversit$.,arbondale, best nown for the production of =ar eting !cales Handboo , has posted its technical reports on the &nternet 1httpCDDwww.siu.eduDdepartmentsDcobaDm tgDosr4. A 7iscuss the use of /LW6&+/6 in scale development and testing. 1""

/LW6&+/6 uses a series of menu.driven screens to guide the mar et researcher through the scale development process. /LW6&+/6 can customize scales for printed questionnaires or for use b$ telephone interviewers at computer screens in a fraction of the time this would ta e without automation. #E AN MR$ AND #E A DM$ &t should be noted that a variet$ of answers are appropriate. +he ones given here are merel$ illustrative. #e %n MR$: Disne& &nformation on movie preferences can be obtained from sources such as httpCDDwww.calstatela.eduDfacult$DsfischoDmedia".html. +here are a number of bodies offering secondar$ data on ratings of individual movies based on surve$s and opinion polls. <rominent among them is =otion <icture >ssociation of >merica 1=<>>4. 6atings for an$ movie can be obtained from their database which can be accessed through the following lin . httpCDDwww.mpaa.orgDmovieratingsDsearchDinde-.htm #e % DM$: Disne& 6atings for similar movies which were released in the same season can be studied to get an idea of possible reaction from audience. >s the movie is being made, audience reaction can be measured using continuous measurement or itemized rating scales. +he movie can be altered to the taste of the audience. 5o simple formula !tud$ responses to various movie t$pes >nal$ze changing trends and tastes :nderstand demographic, ethnic and gender centric preferences :nderstand the popularit$ of actors and actresses and perceptions about them in the minds of the viewers #e % DM$: Die C!'e &dentif$ which attitudes are most relevant to the product and purchase decision. ,ustomers can be segmented along attitudinal and demographic lines such as age, se-, region, social status etc for segmentation. 7esign and position products to satisf$ these segments.

#e %n MR$: Die C!'e Li ert scaleC & li e 7iet ,o e, overall 7iet ,o e offers good value !trongl$ 7isagree 5eutral >gree !trongl$ 7isagree >gree 1 1 2 2 " " # # % %

1"#

7iet ,o e tastes good 7iet ,o e is health$

1 1

2 2

" "

# #

% %

!emantic differential scaleC 7iet ,o e isC 0ood overall 0ood value 0ood tasting Health$ !tapel scaleC 7iet ,o e isC E% E# E" E2 E1 0ood overall .1 .2 ." .# .%

.. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. ..

5ot good overall 5ot good value 5ot good tasting 5ot health$

E% E# E" E2 E1 0ood value .1 .2 ." .# .%

E% E# E" E2 E1 0ood +asting .1 .2 ." .# .%

E% E# E" E2 E1 Health$ .1 .2 ." .# .%

& would recommend the Li ert !cale as it widel$ used to measure attitudes and preferences. #e %n MR$: R!('p!) > Li ert !cale can be used to measure consumer preferences. H$potheses can be formed as to what is important to the customers and what is their preference? +he h$pothesis can be tested against customer opinion using a Li ert scale model surve$. Li ert scaleC !trongl$ 7isagree 5eutral 7isagree & prefer dress shoes to casual shoes 1 2 " >gree # !trongl$ >gree %

#e % DM$: R!('p!) ,onsumer preferences can be used to improve design. =a e tradeoffs between more desirable versus less desirable features. ;ind demographic changes in tastes. :se this information for mar et segmentation. /nsure product.segment fit.

1"%

E*PERIENTIAL LEARNING I+en i,&ing C!-p%)% i.e %n+ N!n/(!-p%)% i.e S(%0es Using Su).e&1$ 1. How man$ t$pes of comparative scales are available in !urve$zO? Answer: Rank order and constant sum are directly available. (The matrix table formatted in spreadsheet form could also be used for paired comparison.) How man$ format options are available for these comparative scales? Answer: One each. How man$ t$pes of noncomparative scales are available in !urve$zO? Answer: emantic differentia! and "ikert scales are directly available. How man$ format options are available? Answer: One each.

2. ". #.

M(C%nn E)i('s!n %n+ A+ P)!-ises 1. !tudents should download and read 11?a4 =c,ann /ric son /-periential Learning file. 2. +he results are presented in 11?c4 >d promises reversed 1!<!! file4. ". a. and b. #ote: The $ronbach alpha values are included in bold italics after each construct%s label below. &ach $ronbach alpha was computed with a sample si'e of ()*. +or the composite variables! each of the ei,ht different composite variables should have values ran,in, from ) to -. >d <romise !cale &tems 1. <roblem removal .728 a. +he brand would help remove a problem one ma$ encounter b. 2ne would go from feeling anno$ance to feeling relief as a result of using the brand. c. :sing the brand would 52+ appeal to oneJs desire for resuming the pursuit of a goal. 2. <roblem avoidance .776 a. :sing the brand would enable one to avoid a potential problem. b. +he brand would dissipate fear and bring one a feeling of rela-ation. c. 2neJs desire for a threat.free pursuit of a goal would 52+ be met b$ using the brand. ". &ncomplete satisfaction .811 a. +he brand would bridge the gap between oneJs e-pectations and the e-isting circumstances in a situation. b. +he brand would bring an optimistic outloo to an otherwise disappointing situation. c. 2neJs desire for more complete satisfaction would 52+ be met b$ using the brand. #. =i-ed approach.avoidance .810 a. +he brand would ta e care of one of the remaining negative aspects of a situation. 1"'

%.

'.

(.

).

b. :sing the brand would avoid impending conflict and bring peace of mind. c. 2neJs desire for more consistenc$ in thoughts about an obBect or issue would 52+ be met b$ using the brand. !ensual gratification .817 a. > sample of the brand would ma e one want more. b. :sing the brand would ta e one from a neutral state to a pleasurable state. c. +he brand would 52+ appeal to oneJs bodil$ senses. &ntellectual stimulation.767 a. +he brand would stimulate oneJs intellect. b. +he brand would relieve boredom. c. +he brand would 52+ appeal to m$ sense of adventure or ris . !ocial approval .904 a. 2ne would be considered more fashionable b$ using the brand. b. +he brand would help one feel less apprehensive in social situations. c. +he brand would 52+ appeal to oneJs desire for social approval. &ntrinsic satisfaction .605 a. :sing the brand is its own reward. b. +he brand would be enBo$ed for its own sa e, not for what it will bring one. c. <ure enBo$ment of the brand would 52+ be the onl$ thing in it for someone.

2HAT 2OULD 3OU DO4 &t should be noted that a variet$ of answers are appropriate. +he ones given here are merel$ illustrative. T5e M%)'e ing Rese%)(5 De(isi!n 1. , 2. &n matching Bobs with applicantJs s ills is critical to the success of =onster. !emantic 7ifferential is an effective scaling technique for capturing varied levels of e-pertise. +he Bob applicants can be required to fill in a profile which includes semantic differential scales for measuring different s ill levels. 2n the other hand the s ill levels required of various Bobs could be captured using the same scales. 8$ using software functionalit$, the Bob descriptions can be automaticall$ matched with applicantJs profiles based on s ills. T5e M%)'e ing M%n%ge-en De(isi!n 1. ,, 7, / 2. +he biggest advantage of =onster from a recruiterJs perspective is efficienc$ at a reduced cost. +his is primaril$ because of a large applicant pool. =onster acts as a centralized database with a ver$ large and varied applicant pool and sophisticated search and matching options. >s a web.based medium, =onster alread$ offers a substantial cost advantage over the conventional means, so further reduction of prices would not be a selling point. +he current search options are ver$ sophisticatedF however, an$ improvement in this front can better the competitive position. +the biggest attraction for emplo$ers is the size and qualit$ of the applicants., so an$ promotional activit$ to attract more resumes will be to =onsterJs advantage. REVIE2 QUESTIONS

1"(

1.

2. ".

#.

%. '. (.

). *.

1?. 11. 12. 1".

> semantic differential is a seven.point rating scale with end points associated with bipolar labels that have semantic meaning. +his scale is used in comparing brand, product, and compan$ images, developing advertising and promotion strategies, new product development studies, and in a variet$ of other applications. +he Li ert scale requires the respondents to indicate a degree of agreement or disagreement with each of a series of statements related to the stimulus obBects. +he differences between the !taple scale and the semantic differential is that in the !taple scale there is no need to pretest the adBectives or phrases to ensure true bipolarit$, and the !taple scale can also be administered over the telephone. +he semantic differential is more popular than the !taple scale. +he maBor decisions involved in constructing an itemized rating scale areC i. +he number of scale categories to use. ii. 8alanced versus unbalanced scale. iii. 2dd or even number of categories. iv. ;orced versus nonforced nature of the scale. v. +he nature and degree of verbal description to emplo$. vi. +he ph$sical form of the scale. +he amount of scale categories that should be used in an itemized rating scale should be between five and nineF however, there is no single, optimal number of categories which would be applicable for all scaling situations. +he difference between balanced versus unbalanced is that a balanced scale has an equal number of favorable and unfavorable categories that are used, whereas the unbalanced scale does not have an equal number of favorable and unfavorable categories. +he decision regarding an odd or even number of categories should depend on whether some of the respondents ma$ be neutral on the response being measured. &f a neutral or indifferent scale response is a possibilit$ for at least some of the respondents, an odd number of categories should be used. &f the researcher wants to force a response to indicate some degree of favorableness or unfavorableness, or it is believed that no neutral or indifferent response e-ists, a rating scale with an even number of categories should be used. +he difference between a forced and nonforced scale is that a forced scale does not have a Gno opinionG or Gno nowledgeG option, whereas a nonforced scale includes a Gno opinionG or Gno nowledgeG categor$. +he nature and degree of verbal description can affect the response to itemized rating scales due to the strength of the adBectives used to anchor the scale. With strong anchors, respondents are less li el$ to use the e-treme scale categories, thus strong anchors result in less variable and more pea ed response distributions. Wea anchors have a tendenc$ to produce uniform or flat distributions. =ulti.item scales consist of a number of rating scale items where the responses are usuall$ summed over the items to determine an overall or total score. >n e-ample would be the Li ert attitude scale given in this chapter. 6eliabilit$ refers to the e-tent to which a scale produces consistent results if repeated measurements are made on the characteristics. +he maBor differences between the two are that test.retest reliabilit$ administers the same scale two different times and measures the correlation between the results. >lternative. forms reliabilit$ uses two equivalent forms of a scale for the two tests. +he internal consistenc$ reliabilit$ approach can be applied to assess the reliabilit$ of a summated scale where several items are summed to form a total score. /ach item can be 1")

1#. 1%.

1'.

1(.

considered to measure the mar eting construct in question and the items should be consistent in what the$ indicate about the construct. 9alidit$ refers to the accurac$ of measurement. 9alidit$ of a scale ma$ be defined as the e-tent to which differences in observed scale scores reflect true differences among obBects on the characteristic being measured, rather than s$stematic or random errors. ,riterion validit$ e-amines whether the measurement scale performs as e-pected in relation to other variables selected as meaningful criteria. ,riterion validit$ is assessed when the data on the scale being evaluated and the criterion variables are collected at the same time. +he relationship between reliabilit$ and validit$ can be understood in terms of the true score model. &f a measure is perfectl$ valid, it is also perfectl$ reliable. &f a measure is perfectl$ reliable, it ma$ or ma$ not be perfectl$ valid, as s$stematic error ma$ still be present. +he lac of reliabilit$ constitutes negative evidence for validit$. 6eliabilit$ is a necessar$, but not a sufficient, condition for validit$. When choosing a particular scaling technique, an attempt should be made to use the scaling technique, which will $ield the highest level of information feasible in the given situation. &n man$ situations it ma$ be desirable to use more than one scaling technique or to obtain additional measures using procedures other than the conventional scaling techniques.

APPLIED PRO#LEMS 1. +he following are e-amples of each scale. !tudent answers ma$ var$, but should have similar form. Li ert !cale &nstructionsC Listed below are statements that describe different opinions about !tore P. <lease indicate how strongl$ $ou agree or disagree with each statement b$ using the following scaleC 1 Q !trongl$ 7isagree 2 Q 7isagree " Q 5either >gree nor 7isagree # Q >gree % Q !trongl$ >gree !trongl$ 7isagree 1. & do most of m$ 1 shopping at store P. 7isagree 5either >gree nor 7isagree " >gree !trongl$ >gree # %

1"*

!emantic 7ifferential &nstructionsC +his stud$ measures different concepts related to !tore P. <lace an P along the line according to how well the subBect is related to the adBectives on the end of the scale. 1. & do most of m$ shopping at store P. 5ever ..C..C..C..C..C..C.. >lwa$s !tapel !cale &nstructionsC <lease evaluate how accuratel$ the word or phrase describes each of the subBects 1e.g., !tore P4 listed below b$ selecting a number for the word1s4. +he more accuratel$ $ou thin the words describe the subBect, the larger the plus number $ou should choose. Rou should select a minus number for words $ou thin do not describe the subBect accuratel$. +he more accuratel$ $ou thin the word describes it, the larger the plus number $ou should choose. +herefore, $ou can select an$ number from E% for words that $ou thin are ver$ accurate, all the wa$ to .%, for words that $ou thin are ver$ inaccurate. ................................................ !tore P ................................................ E% E# E" E2 E1 & do most of m$ shopping at this store .1 .2 ." .# .% =an$ variations of the scale can be constructed. &t should be a multi.item scale and should have anchors that are both understandable and relevant to attitudes on internationalization of the management curriculum. +he scale should contain % to * values and be uniform throughout the items. 6everse scoring should be used. ;or assessing reliabilit$ see the answers to questions 11, 12 and 1". ;or assessing validit$ see the answer to questions 1# and 1%. Listed below are statements, which describe different opinions about the &nternet. <lease indicate how strongl$ $ou agree or disagree with each statementC !trongl$ 7isagree 1#? 5either >gree !trongl$

2.

".

7isagree 1. +he &nternet is a useful source of general information. 2. &t is eas$ to find information on the &nternet. ". When & need information, & would rather read the newspaper than surf the &nternet. 1 2

>gree nor 7isagree " #

>gree %

"

"

#. Li ert scales can be developed to measure the usefulness of ;ord =otor ,ompan$Js Web site at www.ford.com. !cale items should includeC availabilit$ of information, visual search, price information availabilit$, ease of navigation, dealer information and lin ages to dealers and other relevant sites, service to customers, user groups and lin s to user groups, and the visual attractiveness of site. %. 8i.polar adBectives such as reliable . unreliable, on.time deliver$ . late deliver$, trustworth$ . untrustworth$, global service . domestic service, customer friendl$ . customer unfriendl$, competitivel$ priced . high priced, e-cellent service . poor service, etc. can be used. '. !everal different applications can be identified. +he scales should be related to specific mar eting research and corresponding managerial conte-ts. (. +he :6Ls of several mar eting research firms are given in ,hapter 1 1+able 1.14. +he criteria for evaluation should follow the guidelines for rating scale decisions 1see +able 1?.24 ). =an$ t$pes of e-amples can be created. +he Li ert t$pe scales are the most popular. GROUP DISCUSSION 1. +he issues behind this statement center on the appropriateness of a particular scale. +here is a great difference in the results obtained from comparative and noncomparative scales and this distinction should be made. &n addition, for noncomparative scales, the researcher must ma e decisions with respect to the number of scale categories, balanced versus unbalanced scales, odd or even number of categories, forced versus non forced scales, nature and degree of verbal description, and the ph$sical form or configuration. /ach of these decisions will affect the information gathered from respondents and should be discussed. ;inall$, the nature of reliabilit$ is important to consider. /ven though a scale is reliable, it does not mean that the information is pertinent to the issue at hand. =isapplied scales can be ver$ reliable, $et be of little help or even misleading if the$ do not relate to the problem at hand. +he issues of reliabilit$ and validit$ are critical in applied mar eting research, since actual decisions will be made and dollars spent based on the assumed reliabilit$ and validit$ of a scale. 2n the other hand, if the scale has been shown to be reliable and valid from past e-perience, testing reliabilit$ and validit$ is not critical. However, for newl$ devised scales or scales used for the first time in a particular conte-t, reliabilit$ and validit$ should be e-amined to ensure accurate results.

2.

1#1

".

>n$ one of readings can be selected. +he discussion should be related to one or more of the concepts discussed in this chapter.

1#2

S-ar putea să vă placă și