15/03/2021

COPYRIG HT © 2015 PEARSON EDU CATION , IN C.

1-1

Chapter8– Measurement

COPYRIG HT © 2015 PEARSON EDU CATION , IN C.

8-2

1


15/03/2021

LearningObjectives
Afterstudyingthischapter,you shouldbeableto:
ãDescribewhymeasurementandassessmentareimportantto
staffing
ãDescribepatternsindata
ãUnderstandcorrelationandregressionandexplainhoweachis
used
ãDefinebothpracticalandstatisticalsignificance,andexplainwhy
theyareimportant
ãDefinereliabilityandvalidityand explainhowtheyaffectthe
evaluationofameasure
ãExplainwhystandardizationandobjectivity areimportantin
measurement
COPYRIG HT â 2015 PEARSON EDU CATION , IN C.

8-3

WhyIsProper
MeasurementImportant?
Effectivemeasurementanddataanalyticscanresultina
competitiveedge
Improperlyassessing andmeasuringcandidate
characteristicscanleadto:
§ Systematicallyhiringthewrongpeople
§ Offendingandlosinggoodcandidates
§ Exposingyour companytolegalaction

Therearemanylegalissues involvedwithcandidate
assessment andmeasurement

COPYRIG HT © 2015 PEARSON EDU CATION , IN C.

8-4

2


15/03/2021

WhatIsMeasurement?
Measurement istheprocessofassigningnumbersaccordingtosomerule
orconventiontoaspectsofpeople,jobs,jobsuccess,oraspectsofthe
staffingsystem
Themeasures enableimprovementofthestaffingsystembyidentifying

patternsusefulforunderstandingandpredictingrelevantprocessesand
outcomes
Themeasures relevanttostaffingarethosethatassess:
§ Thecharacteristics ofthejob,which enablesthecreationofjobrequirements
andjobrewardsmatrices
§ Aspectsofthestaffingsystem suchasthenumber ofdaysajobpostingisrun,
whereitisrun, andtherecruiting message
Đ Thecharacteristics ofjobcandidates suchasabilityorpersonality
Staffingoutcomes,suchasperformance orturnover

COPYRIG HT â 2015 PEARSON EDU CATION , IN C.

8-5

WhatIsData?
Thenumericaloutcomesofmeasurementaredata
Thereare2typesofdata:
§
§

Predictive data isinformation aboutmeasuresusedtomake
projections aboutoutcomes.
Criterion data isinformation aboutimportant outcomes ofthe
staffingprocess.

o
o

Traditionally, thisdataincludesmeasurementofemployeejobsuccess,
whichistheorganization’suniquedefinitionofsuccessandperformance

inthejobandinthefirm.
Criteriondatashouldalsoincludealloutcomedatathatisrelevanttothe
evaluationoftheeffectivenessofthestaffingsystemagainstitsgoals.
Thismayincludemeasuresofjobsuccess,time-to-hire,promotionrates,
andtenurerates aswellasjobandcompanyengagement, fitwith
companyvalues,andwillingnesstohelpotheremployees.

COPYRIG HT © 2015 PEARSON EDU CATION , IN C.

8-6

3


15/03/2021

TypesofMeasurements
ĐNominal:numbers areassignedtodiscretelabels
orcategories(e.g.,race,gender,collegemajor)
ĐOrdinal:attributesarerankedinascendingor
descending order(e.g.,rankingfrombesttoworst
performance)
ĐInterval:zeropointisarbitrarybutdistance
betweenscoreshasmeaning(e.g.,intelligenceor
interviewscores)
ĐRatio:distancebetweenscoreshasmeaningand
thereisatruezeropoint(e.g.,salary,typingspeed)
COPYRIG HT â 2015 PEARSON EDU CATION , IN C.

8-7


DescribingData
Scoring:Theprocessofassigningnumericalvalues
duringmeasurement
Rawscores:theunadjusted scoresonameasure

§ Criterion-referencedmeasures:measuresinwhichthescores
havemeaninginandofthemselves
§ Norm-referencedmeasures:measuresinwhichthescoreshave
meaningonly incomparisontothescoresofother respondents

Normalcurve:asymmetrical,bell-shaped curve
representingthedistribution ofacharacteristic

COPYRIG HT © 2015 PEARSON EDU CATION , IN C.

8-8

/>
4


15/03/2021

TheNormalCurve

8-9

COPYRIG HT © 2015 PEARSON EDU CATION , IN C.


DescribingtheNormalCurve
Percentilescore:arawscorethathasbeenconverted intoan
expressionofthepercentageofpeoplewhosescorefallsator
belowthatscore
Centraltendency:describesthemidpointorcenterofdata
§ Mean:theaverage ofthescores
§ Median:themiddlescore,orthepointbelowwhich50percentofthescoresfall
§ Mode:themostcommonlyobservedscore(bimodal=twomodes)

Variability:describesthespreadofthedataaroundthemidpoint

§ Range:thedifferencebetween the highest&lowestobservedscore
§ Outlier: scoremuchhigherorlowerthanmostofthescoresinadistribution
Đ Variance:amathematical measure ofspreadbasedonsquareddeviationsofscores
fromthemean
Đ Standarddeviation:positivesquarerootofthevariance;conceptuallysimilartothe
average distancefromthemeanofasetofscores

COPYRIG HT â 2015 PEARSON EDU CATION , IN C.

8-10

5


15/03/2021

StandardScores
Standardscores:Convertedrawscoresthat
indicatewhereaperson’s scoreliesincomparison

toareferentgroup.
§ Acommonstandardscoreisthez score.
§ Az scoreindicates howmanyunitsofstandarddeviationsthe
individual’s scoreisaboveorbelowthemeanofthereferentgroup

Az scoreisnegativewhenthetargetindividual’s
rawscoreisbelowthereferentgroup’s mean,and
positive whenthetargetindividual’s rawscoreis
abovethereferentgroup’s mean

COPYRIG HT © 2015 PEARSON EDU CATION , IN C.

8-11

ConvertingRawScores
toStandardScores

Mean
StdDev

18.25
3.00

Mean 78.25
StdDev 7.46

zscore =(Individual’srawscore– Referent groupmean)/Referentgroup
standarddeviation)
Meaningfullycombiningtherawscoreswouldbedifficult. Combining thez
scoresiseasyandresults inasinglenumber reflecting howeachcandidatedid

onbothoftheassessmentsrelativetotheother candidates.

COPYRIG HT © 2015 PEARSON EDU CATION , IN C.

8-12

6


15/03/2021

Shiftingthe
NormalApplicantTalentCurve
ĐWhenmakingselectiondecisions,itisoftenassumed
thatintheapplicantpool,thedistributionofapplicant
fitwiththejobreflectsthenormalcurve.Alargeburden
isthenplacedontheselection systemtoaccurately
identifywhichcandidatesareinthefarrighttailofthe
normalcurve.
ĐHowever,manyofthemostdesirablepeopleforthe
positionarelikelytobeactivelyandhappilyemployed
elsewhereandaresemi-passivejobseekersatbest.In
thiscase, thedistributionofapplicantfitwiththejob
mightresembletheAdistributionshownonthenext
slide.
8-13

COPYRIG HT â 2015 PEARSON EDU CATION , IN C.

ShiftingtheApplicantTalentCurve


COPYRIG HT © 2015 PEARSON EDU CATION , IN C.

8-14

7


15/03/2021

Shiftingthe
NormalApplicantTalentCurve
ĐIfdonestrategically,sourcingandrecruitingcan
discouragepoorfitsfromapplyingandincreasethe
numberofhighqualitypassiveandsemi-passive
candidateswhoapply.
ĐThisshiftsthecurvetoreflectadistributionlikethat
shownbytheBdistribution.
ĐTheBdistributionclearlyreducestheburdenonthe
selectionsystemtoidentifyqualitycandidatesand
significantlyincreasesthelikelihoodofidentifyinga
high-qualitycandidate.

COPYRIG HT â 2015 PEARSON EDU CATION , IN C.

8-15

CorrelationCoefficient
Correlationcoefficient,alsocalledPearsonsrorthe
bivariatecorrelation,isasinglenumberthatranges

from-1to+1thatreflectsthedirection (positiveor
negative)andmagnitude (strength)oftherelationship
betweentwovariables.

Đ Avalueofr=0indicatesthatvaluesofonemeasureare
unrelatedtovaluesoftheothermeasure.
Đ Avalueofr=+1meansthatthereisaperfectlylinear,
positiverelationshipbetweenthetwomeasures;asvalues
ofonemeasureincrease,valuesoftheother measure
increaseexactlythesameamountinstandarddeviations.
Đ Avalueofr=-1meansthatthereisaperfectlynegativeor
inverserelationshipbetweenthetwomeasures;asvaluesof
onemeasureincrease,valuesoftheothervariabledecrease
exactlythesameamountinstandarddeviations.
COPYRIG HT â 2015 PEARSON EDU CATION , IN C.

8-16

8


15/03/2021

GraphingCorrelations
Scatterplot:graphicalillustrationoftherelationship
betweentwovariables

Đ Eachpoint onthechartcorresponds tohowapersonscoredon
ameasureand howheor sheperformedon thejob


COPYRIG HT â 2015 PEARSON EDU CATION , IN C.

8-17

ScatterPlotofr=-.43
Would this test be useful in making hiring decisions?

COPYRIG HT © 2015 PEARSON EDU CATION , IN C.

8-18

9


15/03/2021

ScatterPlotofaCurvilinear
Relationship(r=.04)

COPYRIG HT © 2015 PEARSON EDU CATION , IN C.

8-19

DiagramsforCorrelations

COPYRIG HT © 2015 PEARSON EDU CATION , IN C.

8-20

10



15/03/2021

DiagramsforCorrelations

COPYRIG HT © 2015 PEARSON EDU CATION , IN C.

8-21

DiagramsforCorrelations

COPYRIG HT © 2015 PEARSON EDU CATION , IN C.

8-22

11


15/03/2021

ExamplesofUsesof
CorrelationCoefficients
ĐRelatingstoresizewithstaffinglevels
ĐRelatingseniorityinafirmwithjobperformance
ĐRelatingthetimetofillajobwithnew-hirequality
ĐRelatingqualityofnewhireswithbusiness
performanceandcustomersatisfaction

COPYRIG HT â 2015 PEARSON EDU CATION , IN C.


8-23

InterpretingCorrelations
ĐSamplingerror: Whenyouusestatistics,including
correlations,todrawinferencesorconclusions,youhaveto
beconcernedaboutsamplingerror.Samplingerroristhe
variabilityinsamplecorrelationsduetochance.
ĐYoucanaddresssamplingerrorthroughstatistical
significancetestingprocedures.

COPYRIG HT â 2015 PEARSON EDU CATION , IN C.

8-24

12


15/03/2021

InterpretingCorrelations,cont.
Statisticalsignificance:thedegreetowhichtheobserved
relationship isnotlikelyduetosamplingerror.

§ Aminimumrequirementforestablishingameaningfulrelationship.

Practicalsignificance:theobserved relationship islarge
enoughtobeofvalueinapracticalsense.

§ Inalargeenough sample,avery smallcorrelationwouldbe

statisticallysignificantbuttherelationshipmaynotbestrong
enough tojustifytheexpenseandtimeofusingthepredictor.
§ Aninexpensiveassessmentsystemmaybeusefulevenifthe
correlationissmall.
§ Alternatively,ifanassessmentthatcorrelated.15withjobsuccess
wasexpensive,tookalongtimetoadminister,andwasnotlikedby
jobcandidates,itmaynotbeworthusingevenifitisastatistically
significantpredictorofjob success.
8-25

COPYRIG HT © 2015 PEARSON EDU CATION , IN C.

MultipleRegression
ĐAstatistical techniquethatpredictsanoutcomeusingone
ormorepredictorvariables;itidentifiestheidealweightsto
assigneachpredictortomaximizethevalidityofasetof
predictors
ĐTheanalysisisbasedoneachpredictor scorrelationwith
theoutcomeandthedegreetowhichthepredictorsare
themselves correlated
ĐMultipleregressionexaminestheeffectofeachpredictor
variableafterstatistically controllingfortheeffectsofother
predictorsintheequation

COPYRIG HT â 2015 PEARSON EDU CATION , IN C.

8-26

13



15/03/2021

XandZCorrelatedwithY
butUncorrelatedwithEachOther

COPYRIG HT © 2015 PEARSON EDU CATION , IN C.

8-27

XandZCorrelatedwithYandHighly
CorrelatedwithEachOther

COPYRIG HT © 2015 PEARSON EDU CATION , IN C.

8-28

14


15/03/2021

MoreTypicalExample

COPYRIG HT © 2015 PEARSON EDU CATION , IN C.

8-29

Exampleofa
MultipleRegressionEquation

Jobsuccesspredicted =Constant+(b1 *Testscore1 )+(b2 *Testscore2 )
+(b3 *Testscore3 )…
Jobsuccesspredicted =10+(2*Interview)+(1*Personality)
+(.2*Jobknowledge)
Ifsomeonescores50ontheinterview, 27onthepersonalitytest, and20onthe
jobknowledgetest,whatisthepredicted jobsuccessscore?
Jobsuccesspredicted =10+(2*50)+(1*27)+(.2*20)
Jobsuccesspredicted =141
141isthencomparedwith predicted jobsuccessscoresofother candidatesto
determine whoshouldbeselected
COPYRIG HT © 2015 PEARSON EDU CATION , IN C.

8-30

15


15/03/2021

WhatIsReliability?
§Reliabilityreferstohowdependable orconsistent a
measureisinassessing aparticularcharacteristic.
§Measurementerrorinfluences reliability.
§Measurementerrorcanberandomorsystematic.
§Toevaluateameasure’sreliability,you should consider:
§ Thetypeofmeasure
§ Thetypeofreliabilityestimatereported
§ Thecontextinwhichthemeasurewillbeused

COPYRIG HT © 2015 PEARSON EDU CATION , IN C.


8-31

ReasonsforDifferingScores
onaTestorAssessment
Allofthesefactors,aswellasothers,caninfluence
reliability.Thatiswhytestsorassessment tools
should bestandardizedintheiruse.
Đ Temporaryphysicalor psychologicalstate
Đ Environmentalfactors
Đ Version,orform,ofthemeasure
Đ Differentevaluators

COPYRIG HT â 2015 PEARSON EDU CATION , IN C.

8-32

16


15/03/2021

TypesofErrors
ĐRandomerror:errorthatisnotduetoany
consistentcause
ĐSystematicerror:errorthatoccursbecauseof
consistentandpredictablefactors
ĐDeficiencyerror:errorthatoccurswhenyoufailto
measureimportantaspectsoftheattributeyou
wouldliketomeasure

ĐContaminationerror:errorthatoccurswhenother
factorsunrelatedtowhateverisbeingassessed
affecttheobservedscores
COPYRIG HT â 2015 PEARSON EDU CATION , IN C.

8-33

Deficiency,Contamination,and
Relevance

COPYRIG HT © 2015 PEARSON EDU CATION , IN C.

8-34

17


15/03/2021

InterpretingReliabilityCoefficients

The proper interpretation of reliability coefficients depends on
the type of reliability being assessed and the purpose of the
measure.

COPYRIG HT © 2015 PEARSON EDU CATION , IN C.

8-35

TypesofReliability

§Test-retestreliability reflectstherepeatabilityofscoresover
timeandthestabilityoftheunderlyingconstruct being
measured
§Alternateorparallelformreliability indicateshow
consistentscoresarelikelytobeifapersoncompletestwo
ormoreformsofthesamemeasure
ĐInternalconsistencyreliability indicatestheextenttowhich
itemsonagivenmeasureassessthesameconstruct
ĐInter-raterreliability indicateshowconsistentscoresare
likelytobeiftheresponsesarescoredbytwoor more
ratersusingthesameitem,scale,orinstrument

COPYRIG HT â 2015 PEARSON EDU CATION , IN C.

8-36

18


15/03/2021

StandardErrorofMeasurement
Thestandarderrorofmeasurement (SEM)isthemarginof
errorthatyou shouldexpectinanindividualscorebecause
oftheimperfectreliabilityofthemeasure.Itrepresentsthe
spreadofscoresyoumighthaveobserved hadyou tested
thesamepersonrepeatedly.
Theconfidenceintervalrepresentsthedegreeof
confidencethataperson’s “true”scorelieswithintheir
earnedscoreplusorminustheSEM,givensomelevelof

desiredconfidence.
Thelowerthestandarderror,themoreaccuratethe
measurements.
§ IftheSEMis0,theneachobservedscoreisthatperson’s true score

COPYRIG HT © 2015 PEARSON EDU CATION , IN C.

8-37

WhatIsValidity?
Validityreferstohowwellameasureassessesagiven
constructandthedegreetowhichyoucanmakespecific
conclusionsorpredictions basedon observed scores.

§ Validitycantellyouwhatyoumayconcludeorpredict about
someonebasedonhisorherscoreonameasure,thusindicating the
measure’susefulness.
§ Validity willtellyouhowusefulameasureisforaparticular
situation; reliability willtellyouhowconsistent scoresfromthat
measurewillbe.
§ Youcannotdrawvalidconclusions unlessyouaresurethatthe
measureisreliable.Evenwhenameasureisreliable, itmaynotbe
valid.
Đ Youmightbeabletomeasureapersons shoesizereliablybut itmay
notbeusefulasapredictor ofjobperformance.

Anymeasureusedinstaffingneedstobebothreliableand
validforthesituation.
COPYRIG HT â 2015 PEARSON EDU CATION , IN C.


8-38

19


15/03/2021

ReliabilityandValidity

COPYRIG HT © 2015 PEARSON EDU CATION , IN C.

8-39

WhatIsValidation?
Validationisthecumulativeandongoingprocess of
establishingthejobrelatednessofameasure
Therearethreetypesofvalidationprocesses:

Đ Content-relatedvalidation:Demonstratingthatthecontentofa
measureassessesimportantjob-relatedbehaviors
Đ Construct-relatedvalidation: Demonstratingthatameasure
assessestheconstruct,orcharacteristic,itclaimstomeasure
Đ Criterion-relatedvalidation:Demonstratingthatthereisa
statisticalrelationshipbetweenscoresfromameasureandthe
criterion,usuallysomeaspectofjobsuccess

COPYRIG HT â 2015 PEARSON EDU CATION , IN C.

8-40


20