surveys in software engineering
TRANSCRIPT
SurveysinSoftwareEngineering
Rafael Maiani de Mello
Marco Torchiano
Daniel Méndez Guilherme H. TravassosFurther coaches
Ground rules for today
1. Whenever you have a question/remark: share them with the group!
2. Stay flexible
3. There are no further rules J
Feelfreeto…
copy,share,andalter,
filmandphotograph(aslongaswelookgreat)
tweetandliveblog
today’spresentationsgiventhatyouattributeittoitsauthor(s)andrespectitsrightandlicensesofitsparts.
Whoare we?
MarcoTorchianoPolitecnico diTorinohttp://softeng.polito.it/torchiano/
RafaelMaiani deMelloPontificalCatholicUniversityofRiodeJaneirohttp://inf.puc-rio.br/~rmaiani
GuilhermeHorta TravassosCOPPE,FederalUniversityofRiodeJaneirohttp://www.cos.ufrj.br/~ght
DanielMéndezTechnicalUniversityofMunichhttp://www4.in.tum.de/~mendezfe/
Whoareyou?
Quick round:
• Who are you?
• What is your experience in conducting survey research?
• What are your expectations?
What doyou think?
Whydoweneedsurveyresearchinsoftwareengineering?
AgendaTime Topic
09:00– 11:00 SessionI- IntroductiontosurveysWherewewillprovidethebasictheoreticalconceptsofpopulationsurveys:generalmethod,sourceoferrors,sampling,instrumentdesign.
11:00– 11:30 Morningbreak
11:30– 13:00 SessionII- BestpracticesWherewewillfocusonthekeyaspectsofdesigningandconductingsoftwareengineeringsurveysandpresentobservedissuesandevidencebasedlessons.
13:00– 14:30 Lunch14:30– 16:30 SessionIII- Hands-on(BYOL)
Wheretheparticipantsareexpectedtodesignandimplementasimplesurveyonarealonlinetool.BringYourOwnLaptop,ortabletatleast.
16:30– 17:00 Afternoonbreak
17:00– 18:30 SessionIV– Q&AWherewewilldiscusswiththeparticipantsaboutthemostimportantissuesandcomeupwithsomegeneralrecommendation.
19:30– 22:00 TourCiudadRealReceptionattheTownHallofCiudadRealatCasa-Museo López-Villaseñor
SessionI
IntroductiontoSurveyResearch
Session I
Introduction to Survey research
BigPicture…1st layer
9
Philosophy of science
Principle ways of working
Methods and Strategies
Fundamental tools
Epistemology
Empirical methods
Statistics
Hypothesis testingCase studies
Logic
Examples
Theories
ESEreliesoneverylayer!
10
Philosophy of science
Principle ways of working
Methods and Strategies
Fundamental tools
Setting of Empirical Software Engineering:
§ Theory building and evaluation
are supported by
§ Methods and Strategies
Analogy: Theoretical and Experimental Physics
BigPicture…2nd layer
11
Theory/System of theories
(Tentative)Hypotheses
Observations / Evaluations
Study Population
Induction
PatternBuilding
Deduction
Falsification /Support
TheoryBuilding
BigPicture3rd layer:Methods
12
• EachmethodIcanapply…• Hasaspecificpurpose
• Reliesonaspecificdatatype
Purposes
• Exploratory• Descriptive• Explanatory
• Improving
DataTypes
• Qualitative• Quantitative
Example: Grounded Theory
(Tentative)Hypotheses
Study Population
Qualitative Data
DescriptiveExploratory, or
Explanatory
Further reading: Vessey et alA unified classification system for research in the computing disciplines
Theory/System oftheories
(Tentative)Hypotheses
Observations / Evaluations
Study Population
PatternBuilding
Falsification /Support
TheoryBuilding
BigPicture3rd layer:MethodsFormal /
ConceptualAnalysis
Grounded Theory
Confirmatory• Case & Field
Studies• Experiments • Simulations
Survey andInterview Research
• EthnographicStudies
• Folklore Gathering
Exploratory• Case & Field
Studies• Data Analysis
ObservationalStudies
Survey(Cross-Sectional) Case study Case-Control
Survey
Systematicobservationalmethodtogatherqualitativeand/orquantitativedatafrom(asampleof)entitiesto
characterizeinformation,attitudesand/orbehaviorsfromdifferentgroupsofsubjectsregardinganobjectofstudy
15
Descriptivestatistics+Analyticstatistics
Planning
Surveyprocess
16
Define research objectives
Chose collection mode Chose sampling frame
Questionnaireconstruction and pretest
Design and select sample
Recruit and measure
Data coding and editing
AnalysisPost-survey adjustments
Surveyprocess
17
Define research objectives
Chose collection mode Chose sampling frame
Questionnaireconstruction and pretest
Design and select sample
Recruit and measure
Data coding and editing
AnalysisPost-survey adjustments
Research question +Characterization of constructs and population
What is the productivity ofJava software developers?
Parameter/ Construct
Target population
Surveyprocess
18
Define research objectives
Chose collection mode Chose sampling frame
Questionnaireconstruction and pretest
Design and select sample
Recruit and measure
Data coding and editing
AnalysisPost-survey adjustments
How many lines of Java code have you written in the last week?
Measurement
Surveyprocess
19
Define research objectives
Chose collection mode Chose sampling frame
Questionnaireconstruction and pretest
Design and select sample
Recruit and measure
Data coding and editing
AnalysisPost-survey adjustments
A couple hundreds
Response
Surveyprocess
20
Define research objectives
Chose collection mode Chose sampling frame
Questionnaireconstruction and pretest
Design and select sample
Recruit and measure
Data coding and editing
AnalysisPost-survey adjustments
200 LOCs
Edited response
Surveyprocess
21
Define research objectives
Chose collection mode Chose sampling frame
Questionnaireconstruction and pretest
Design and select sample
Recruit and measure
Data coding and editing
AnalysisPost-survey adjustments
Research question +Characterization of constructs and population
What is the productivity ofJava software developers?
Parameter/ Construct
Target population
Surveyprocess
22
Define research objectives
Chose collection mode Chose sampling frame
Questionnaireconstruction and pretest
Design and select sample
Recruit and measure
Data coding and editing
AnalysisPost-survey adjustments
Developers working for companies in the region having NACE activity code J
62.0.x
Frame population
Surveyprocess
23
Define research objectives
Chose collection mode Chose sampling frame
Questionnaireconstruction and pretest Design and select sample
Recruit and measure
Data coding and editing
AnalysisPost-survey adjustments
2 developers in each of 100 randomly selected companies
Sample
Surveyprocess
24
Define research objectives
Chose collection mode Chose sampling frame
Questionnaireconstruction and pretest
Design and select sample
Recruit and measure
Data coding and editing
AnalysisPost-survey adjustments
• [email protected]• [email protected]…
Respondents
Measurementvs Representation
25
Define research objectives
Chose collection mode Chose sampling frame
Questionnaireconstruction and pretest Design and select sample
Recruit and measure
Data coding and editing
AnalysisMake adjustments
Construct Target population
Measurement
Response
Edited response
Frame population
Sample
Respondents
Post-survey adjustments
Instrument
Measurementperspective
• Construct
• Measurement
• Response
• Editedresponse
26
Define research objectives
Chose collection mode Chose sampling frame
Questionnaireconstruction and pretest Design and select sample
Recruit and measure
Data coding and editing
AnalysisPost-survey adjustments
μi
Yi
yiyip
Construct
• Elementofinformationsoughtbyresearchers• Examples
• Howmanynewjobscreated• Howmanyincidentsofcrimewithvictims
• Whichdevelopmentstoolsused• Formulation
• Easytounderstand
• Imprecise• Abstract
27
μi
Construct
• LevelofAbstraction/Measurementperspective
• Directlyobservable
• E.g.Staffforaproject
• Afewdefinedwaystomeasure
• Nondirectlyobservable
• Intentiontoadoptatechnology
• Nosinglewell-definedmeasure
28
Measurement
• Howtogatherinformationaboutconstructs• Objectivemeasures
• Electronic• Physical
• Answerstoquestions• Visual(formalquestionnaires)• Oral(structuredorsemi-structuredinterviews)
29
Yi
Validity
• Gapbetweenconstructsandmeasurement• Ideallythemeasureistheresultofjustoneamongseveral
possibletrials• Inpracticethemeasurementmayintroduceanerror
• Eachtrialintroducesadifferenterror
• Validity:=correlationbetweenYandμ
30
Yit = µi + ✏it
Response
• Theactualdatacollectedthroughthesurvey• Aquestionmayrequire
• Searchownmemory• Accessrecords
• Askotherpersons• Closedquestionsalreadycontainpossibleanswers• Sometimesaresponseisnotprovided
31
yi
• Gapbetweentheidealmeasurementoutcomeandresponseobtained
• Responsebias• Systematicmisreporting
• Reliability
• Variabilityoverseveraltrials
Measurementerror
32
yi � Yi
Editedresponse
• Reviewprocessbeforeusingdata• Rangechecks
• Consistencychecks• Illegibleanswersdetection
• Skippedquestions• Outlierdetection
33
yip
Processingerror
• Gapbetweenvariablesusedinanalysisandthoseprovidedbytherespondent
• Erroneousoutlieridentification• Codingerror
34
Errors– measurementpov
35
Construct
Measurement
Response
EditedResponseProcessing
Error
Measurement Error
Validity
Representationperspective
36
• Targetpopulation
• Framepopulation
• Sample
• Respondents
• Post-surveyadjustments
Define research objectives
Chose collection mode Chose sampling frame
Questionnaireconstruction and pretest Design and select sample
Recruit and measure
Data coding and editing
AnalysisPost-survey adjustments
Y
YC
ya
yr
Representationperspective
37
Target
Frame
Sample
Respondents
Targetpopulation
• Thesetofunitstobestudied• Abstractpopulationdefinition
• E.g.softwareprojects• Time?
• Insoftwarecompaniesonly?• Italiancompaniesonly?• Completedorjuststartedprojects?
38
Framepopulation
• Allunitsinthesamplingframeconstitutetheframepopulation• Intheory
• Thesubsetoftargetpopulationthathasachancetobeselected
• Inpractice• asetofunitsimperfectlylinkedtothetargetpopulation
members
• E.g.telephonenumbers
39
Framinginstrument• The(conceptual)instrumentusedtoidentifytheunitsofstudy
• Householdphonenumberstogetpersons
• Companyrecordstogetemployees• CustomerIDstogetcustomers
• SocialNetworkIDstogetmembers• Warning:oftentheframeelementsareindirectlylinkedtothe
unitsofanalysis,throughrespondents.E.g.
• UoA:softwareprojects(UoA)• Respondents:developers
• Frameelements:softwarecompanies
40
Target
Frame
CoverageError
41
Undercoverage
Ineligible units
Covered population
Realitycheck
• During2012USAPresidentialElectionsCampaignbecauseofaneffectofFederalregulationspollingcellphoneswasmoreexpensive
• Asaresult,manypublicpollsleavecellphoneusersoutoftheirsamples
• Duetothegrowingpopularityofcellphonesastheonlypointofcontactforyoungvotersandminorities,poolersleftkeyconstituenciesforObamaoutofthepollsandskewedthenumbersforRomneyinsomesamples
42http://www.politico.com/news/stories/1112/84103.html?hp=l1
“That’swhysomepolls lookedsodifficultforthepresident,becausetheywereunder-polling theelectorateforthepresident”
J.Messina(CampaignManagerforObama)
Coveragebias
• Twofactors1. Differencebetweencoveredandnotcoveredpopulation
• Y:meanoftarget• YC:meanofcovered YU:meanofuncovered
2. Proportionofnoncoveredpopulation• C:#coveredunitsU:#uncoveredunits
43
Y C � Y =U
C(Y C � Y U )
Sample
• Unitsselectedfromtheframepopulation• Timeandcostopportunity
• Sampling:==Deliberatenon-observation• Mayintroducedeviationbetween
• Samplestatistic• Fullframestatistic
44
SamplingDesign
• Strategyfollowedtoestablishthesamplefromtheframepopulation
• Non-probabilisticsamplingdesignsinclude:1. Accidentalsampling(simplyuseconvenience)
2. Judgementsampling(applysometechnicalcriteriatosample)
3. Snowballing(sharethesamplingdecisionwithpartofthesubjects)
4. QuotaSampling(establishfixedquotabygroups)
45
SamplingDesign
• Probabilisticsamplingdesignsinclude:1. Simplerandomsampling(randomselect"n"unitsfromthe
framepopulation)2. StratifiedSampling(simplerandomsamplingfromeach
stratumestablishedintheframepopulation)• Example:Tosamplejavadevelopersbycountry
46
SampleSizeFormula
• Recommendedwhenworkingwithprobabilisticsamplingdesigns
• SS:samplesize
• Z: Z-value,establishedthroughaspecifictable(Z=2.58for99%ofconfidencelevel,Z=1.96for95%ofconfidencelevel
• p: percentage selecting a choice, expressed as decimal (0.5 used as default forcalculating sample size, since it represents the worst case).
• c:desiredconfidenceInterval,expressedindecimalpoints(Ex.:0.04).
47
SampleSizeFormula
• Correctionformulabasedonafinitepopulationwithapopsize
48
Population Confidence LevelConfidence
IntervalSample Size
10,000 95% 0.01 4,899
10,000 95% 0.05 370
500 95% 0.01 475
500 95% 0.05 217
Samplingerror
• Samplingbias• Systematicexclusionofsomemembers
• Orsignificantlyreducedchanceofselection• Samplingvariance
• Idealsetofsamplesalldrawnfromthesameframe
49
Vs =
SX
s=1
�ys � Y C
�2
S
Samplingerrorreduction
• Probabilisticsamplingü Allunitshavenonzeroselectionprobability
• Stratifiedsamplingü Representationofkeysub-populationsiscontrolled
• Elementsamplesü Asopposedtoclustersamples
• Samplesize
50
Respondents
• Thesubsetofsampleforwhichameasurementcouldbecollected
ü Itemmissingdata:incompletemeasures• Fullparticipation(i.e.100%responserate)realisticallypossible
onlyforinanimateunits
51
Respondents
52
000000000000000000000000000000000000000000
111111x1111111111111111111111111111x11111111111111x111111111x111111111111111111111111111xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
Framedata Interview data
Data items
Samplecases
Respondents
Nonrespondents
Item missing data
Non-responseerror
• Non-responsebiasü Non-responserate:mS /nSü Differencebetweenrespondentsandnon-respondents
53
yr � ys =ms
ns(yr � ym)
Post-surveyadjustment
• Weightingü Compensateunder-representationdueto
ü Nonresponsepatternsü Mismatchbetweenframeandtargetpopulation
• Imputationü Itemmissingdataarereplacedbyestimations
54
Session II
Best Practices on Planning Surveys
DisclaimerThere is no universal silver bullet!
BestPractices
Defining research objectives
Sampling
Questionnaire Design
Recruiting
Characterizing the Target Population
Designresearchobjectives
Challenge:
Knowthelimitationsofsurveyresearch
à Surveyresearchoptsforanswersthatrelyonexperiences,opinions,andobservations(folklore)oftherespondents
• Developinternalquestionstohelpyoudepictingtheresearchobjective
• Optfordescriptivequestions(“whatishappening?”)orexplanatoryquestions(“whyisthishappening?”)ratherthannormativequestions(“whatshouldwedo?”)
Designresearchobjectives
Challenge:
Identifytherealtargetpopulation
à Avoidtorestrictthetargetpopulationbasedonfactorssuchasitssizeoritsavailability.
• Basedontheresearchobjectives,answerthefollowingquestion:“Whocanbestprovideyouwiththeinformationyouneed?”,insteadofanswering“Whoareprobablyavailabletoparticipate?"
BestPractices
Defining research objectives
Sampling
Questionnaire Design
Recruiting
Characterizing the Target Population
CharacterizingtheTargetPopulation
Challenge:
Identifyingthesurveyunitofanalysis
à Thesurveysubjects(individuals)aretypicallythesurveyunitofanalysis.However,insomecasesitmaybeagroupofindividuals,suchashouseholdsororganizationalunits/projectteamsinSEresearch
• Basedontheresearchobjective,identifywhichentityshouldbeusedtoguidesamplinganddataanalysing• Forinstance,investigating Javadevelopers programmingpracticeisaresearch
objectivedifferentfrominvestigatingjavaprogrammingpracticeinsoftwarehouses
Unit of analysis
CharacterizingtheTargetPopulation
Image source: http://www.telegraph.co.uk/finance/personalfinance/8080294/Poorest-households-hit-15-times-harder-by-Government-cuts.html
How do developers perform code debugging?
How code debugging have been performed by developers from software houses?
List of software developers
List ofsoftwarehouses
Challenge:
CharacterizingthesubjectsandunitsofanalysisinSEsurveys
à Differentresearchobjectivesmaydemanddifferentattributestocharacterizingindividuals/groupsofindividualsinvolvedinthesurveys
à Whatattributesarenecessarytoidentifya“representative”population?
• Standardscanbeespeciallyhelpfultoprovidescalesandevennominalvalues• Forinstance,CMMI-DEVmaturitylevelcanbeusedtocharacterize
organizationalunitsregardingtheirmaturityinsoftwareprocess.RUProlescanbeusedtocharacterizesubjects’currentposition
CharacterizingtheTargetPopulation
CharacterizingtheTargetPopulation
Challenge:
CharacterizingthesubjectsandunitsofanalysisinSEsurveys
• Individuals canbecharacterizedthroughattributessuchas:experienceintheresearchcontext,experienceinSE,currentprofessionalrole,locationandhigheracademicdegree]
• Organizations canbecharacterizedthroughattributessuchas:size(scaletypicallybasedinthenumberofemployees),industrysegment (softwarefactory,avionics,finance,health, telecommunications, etc.),locationandorganization type(government, privatecompany,university,etc.)
• Projectteams canbecharacterizedthroughattributessuchasprojectsize;teamsize,client/productdomain(avionics, finance,health,telecommunications, etc.)andphysicaldistribution
BestPractices
Defining research objectives
Sampling
Questionnaire Design
Recruiting
Characterizing the Target Population
Sampling
Challenge:LookingfortheFramePopulation
à SuitablesamplingframesarerarelyavailableinSEresearch.Weoftenneedtoperform“indirectsampling”(forinstance,thereisnoyellowpagesforsoftwareprojectsinacountry).
• Firstofall,youshouldsearchforcandidatesofsourcesofpopulation.Avoidtheconvenienceonsearchingcandidates,tryingtoanswer:“Wherearepresentativepopulationfromthesurveytargetpopulationorevenalltargetpopulationisavailable?”
Image source: http://www.bryan-allen.com/Photography/General/i-fsVh3h2
Sampling
Anuniverseofimperfectalternatives!
SamplingAgoodsourceofpopulation...
• ...should not intentionally represent a segregated subset from the targetpopulation, i.e., for a target population audience “X”, it is not adequate tosearch for units from a source intentionally designed to compose a specificsubset of “X”
• ...should not present any bias on including on its database preferentially onlysubsets from the target population. Unequal criteria for including search unitsmean unequal sampling opportunities
• … allow identifying all source of population’ units by a distinct logical ornumerical id
• ...should allow accessing all its units. If there are hidden elements, it is notpossible to contextualize the population
Sampling
Challenge:
LookingfortheFramePopulation
• InthecaseofsurveyshavingSEresearchersastargetpopulation,youcanuseresultsfrompreviouslyconductedsystematicliteraturereviews(SLR)regardingyourresearchtheme• SocialnetworksaddressedtointegrateacademicssuchasResearchGate and
Academia.edu canbealsousefulinthiscontext.
SamplingChallenge:
LookingfortheFramePopulation
• Lookforcataloguesprovidedbyrecognizedinstitutes/associations/governmentstoretrieverelevantsetofSEprofessionals/organizations.Someexamples:
• SEI(www.sei.cmu.edu) instituteprovidesanopenlistoforganizationsandorganizationalunitscertifiedineachCMMI-DEVlevel.
• FIPA(www.tivia.fi/in-english)providesinformationregardingFinlandITorganizationsanditsprofessionals.
• CAPES(www.capes.gov.br/) providesatoolforaccessinginformationregardingBrazilianresearchgroups.
Sampling
Challenge:
LookingfortheFramePopulation
• Sourcesavailable inthewebsuchasdiscussion groups,projectsrepositories andworldwideprofessional socialnetworks canbehelpfultoidentifyrepresentativepopulationscomposedbySEprofessionals
Such sources can restrict at any moment the access to the
content available!
Sampling
Challenge:
LookingfortheFramePopulation
• Howtofindtheframepopulationinthesourceofpopulation?• Onceyouhaveidentifiedasourceofpopulation,youneedtoestablish
steps/procedurestosystematicallydepictingthesurveysamplingframe.• Suchpracticeisimportanttoassessthesamplesrepresentativeness,
alsosupportingfuturere-executions
Sampling
Challenge:
Establishingthesurveysamplesize
• ParticipationratesinvoluntarysurveysinSEperformedoverrandomsamplestendtobesmall(lowerthan10%).Whataretheimplicationsonresponserates,butalsoonrepresentativeness?
• Takepreferencetoprobabilisticsamplingdesigns.• Independentfromtheamountofrespondents,itwillbepossibletocalculate
theresultsconfidence
• Involuntarysurveyswithpractitioners,establishsignificantlyhighersamplesizes,consideringtheexpectationofaverylowparticipationrate
Image source:http://www.playbuzz.com/viralpx/a-what-do-people-really-think-about-you
BestPractices
Defining research objectives
Sampling
Questionnaire Design
Recruiting
Characterizing the Target Population
QuestionnaireDesign
Challenge:Todesignaclear,simpleandconsistentsurveyquestionnaire
Remember:
Bad questionnaires can led subjects initially willing to participate to give up!
QuestionnaireDesign
Challenge:
Todesignaclear,simpleandconsistentsurveyquestionnaire
• Usesimpleandappropriatewordingforthesurveyquestions• Avoidtechnicaltermsasmuchaspossibleordefinetheminthe
questionnaire,accordingtothesurveytargetpopulation
• Takepreferencetodesignshortquestionsregardingasingleconcept• Avoiddoublebarreledquestions
• Avoidvaguesentenceswhilewritingsurveyquestions
Image source: http://www.fooj.it/wp-content/uploads/2015/07/boring-office.jpg
QuestionnaireDesign
Inyouropinion,doyouagreeordisagreethatcoderefactoringisaneed?Andwhataboutcodesmelldetection?
a) Istronglyagreeb) Ipartiallyagreec) Iagreed) Idisagree
QuestionnaireDesign
Coderefactoringisanessentialpracticeforimprovingtheunderstandingofobject-orientedcode.
a) Totallyagree
b) Partiallyagree
c) Neitheragreenordisagree
d) Partiallydisagree
e) Totallydisagree
QuestionnaireDesign
Challenge:
Todesignaclear,simpleandconsistentsurveyquestionnaire
• Avoidbiased questions,whichcanbedonebycarefullyphrasingthequestionsthatdonotsuggestlikelyanswersorresponses
• Avoidingsensitive questions• InSEcontext,thesensitivequestionscanbeaboutrespondentsincome,
opinionaboutorganizationormanagement,etc.
• Avoidtoaskaboutfarpastevents
QuestionnaireDesign
Doyoupreferworkinginprojectsfollowingagilemethodsorthosefollowingusualnon-agile approaches?
Considering themaincharacteristicsofthelast10softwareprojectsyouhaveworkedon,pleaseanswerthefollowingquestions:
Askingage,gender,maritalstatusforcharacterizingrequirementsengineers
QuestionnaireDesign
Challenge:
Todesignaclear,simpleandconsistentsurveyquestionnaire
• Itisimportanttoavoiddemandingquestions(requiringtoomucheffortfromrespondentstoanswer)
• Avoiddoublenegatives
QuestionnaireDesign
Afterreadingtheattachedpapersregardingnonfunctionalrequirements(NFR),pleaseanswerthefollowingquestions:
1. WhichofthefollowingNFRdoyoudisagreearenotrelevantinthecontextofreal-timesystems?
…
QuestionnaireDesign
Challenge:
Todesignaclear,simpleandconsistentsurveyquestionnaire
• Becarefulonselectingthe ResponseFormat!• Wrongchoicesofresponseformatmayleadyouto:
• Losepreciousdata• Losetheopportunityofapplyingrelevantstatisticaltests
• Significantly(andunnecessarily)increasedataanalysisefforts
QuestionnaireDesign
Free-text
Numeric values
• Open questions• Allow coding• Content analysis• High effort on data
analysis
• Open questions• Allow a wide range
of statistical analysisInterval
Scale
• Closed questions• Not necessarily equally
distributed intervals• Significantly restricts
statistical analysis
Ordinal/ Likert scale
• Closed questions• Intervals are
considered equally distributed
• Statistical analysis is less restrictive than Interval Scale
Nominal• Closed questions• Statistical analysis
based on frequency
QuestionnaireDesign
HowmuchexperiencedoyouhaveinJavaprogramming?
a) VeryHighexperienceb) HighExperiencec) FewExperienced) VeryFewexperience
HowmuchexperiencedoyouhaveinJavaProgramming?
a) Lessthanoneyearb) 1yearto3yearsc) 3yearsto5yearsd) Morethan5years
HowmuchexperiencedoyouhaveinJavaprogramming?
__5__years
HowmuchexperiencedoyouhaveinJavaprogramming?
I have been working with Java programming atcompanies since 2011. Before, I got my firstJava certification in 2009, when I startedworking in personal projects. But I havedifficult with object-oriented parts…_________
DoyouhaveexperienceinJavaprogramming?
()Yes()No
BestPractices
Defining research objectives
Sampling
Questionnaire Design
Recruiting
Characterizing the Target Population
RecruitingChallenge:
Controllingrecruitmentandparticipation
• Sendindividualbutstandardinvitationmessages• Itisexpectedthatgreatmostoftheindividualmessagessentwillberead
• Avoid"spreadingspree":mailinglists,foruminvitationmessages,crowdsourcingtools(suchasAmazonMechanicalTurk)• Youwillhavefewornocontrolonwhoreadtheinvitation.So,whowas
effectivelyrecruited?
• Neverallowforwarding(whichisdifferentfromsnowballing)!• Itwillviolatethesample
• Sendaquestionnaire’sindividualtokentoeachsubject
RecruitingChallenge:
Stimulatingparticipation
• Remindersshouldbeusedwithcare.• Avoidremindingwhoalreadyhadparticipated
• Avoidremindingmorethanonce
• Theinvitationmessageshouldclearlycharacterizetheinvolvedresearchers,theresearchcontextandpresenttherecruitmentparameters
• Includeintheinvitationmessageacomplimentandanobservationregardingtherelevanceofsubjectparticipation
Image source: http://quotesgram.com/you-are-important-to-me-quotes/
Challenge:
Stimulatingparticipation
• Establishafiniteandnotlongperiodtoanswerthesurvey• One-twoweeks
• Offerrewards(raffles,donations,payments,sharingresults)• Takeintoaccountthelocalpolicies
Image source: http://quotesgram.com/you-are-important-to-me-quotes/
Recruiting
BestPractices
Defining research objectives
Sampling
Questionnaire Design
Recruiting
Characterizing the Target Population
PilotingtheSurveyChallenge:
Youhaveonlyoneshot!Onceyoustartedthesurvey,thereisusuallynowayback
• Pilotthepopulationandsamplingactivitiesü Usea(smaller)sampleofthesamplingframe,reproducingallplannedstepsü Willallowyoutochecktheadequacyoftheframepopulationtoyoursurvey.
• Pilotthequestionnaireü Isitclear,unambiguous,didyoumaybemisssomequestions?
• Pilottherecruitmentü Doesitisworkingeffectively?
• Pilotthedataanalysisü Doyouhaveplannedfortheproperdataanalysistechniques?Whatisthenecessary
dataquantityand quality?
SessionIII
Hands-On
Session III
Hands-On!
The“Plan”1. Defineupto4researchobjectives2. Teamassignment
<lunch>
1. ShortintroductionintothetoolFor each team:2. (Online)Surveydesignand implementation3. Piloting /Testing4. Wrap-up5. (Optional:Running survey with ESEMparticipants)
1. Your research objective2. Your (coarse) research questions3. Survey target population4. Survey unit of analysis5. Possible sources of population
Definitionof...
à Everyone sketcheshis/heridea for asurveyonasticky note
à Wemakea(plenary)selection and teamassingment
94
15-20 minutes
5 minutes
Break
Back-UpQuestions• Are functional and non-functional requirements really
distinct?• Can current software testing techniques support the test
of context awareness systems?• What development practices can contribute more to
decrease the software technical debt?• What could be the software engineering gap when
developing scientific (e-science) software?• What is the impact of Continuous Software Engineering
on software productivity and quality?• What is software?
The“Plan”1. Defineupto4researchobjectives2. Teamassignment
<lunch>
1. ShortintroductionintothetoolFor each team:2. (Online)Surveydesignand implementation3. Piloting4. Wrap-up5. (Optional:Running survey with ESEMparticipants)
Login
http://ww2.unipark.de/www/
Login
http://ww2.unipark.de/www/
Team1User:iasese_1Password:A2C04aq.
Team2User:iasese_2Password:A2C04sw.
Team3User:iasese_3Password:A2C04de.
Team4User:iasese_4Password:A2C04fr.
Selectyourproject
• Pleasesticktoyourownproject(watchtheteamnumber)
• Pleasetakecarewhatyoudo,youhaveadminrightsinthesystem
Startediting
Participationlink(page-based)questionnaireeditor
Dataexport
Don’ttouchJ
Questionnaire
ExportingdataDataexport(takesabit)
Codebook
BeforeyoustartTestandreset
Acclimatizationphase
Team1User:iasese_1Password:A2C04aq.
Team2User:iasese_2Password:A2C04sw.
à Everyteamgets to know the toolà Incaseofquestions,askJ
Team3User:iasese_3Password:A2C04de.
Team4User:iasese_4Password:A2C04fr.
10 minutes
http://ww2.unipark.de/www/
The“Plan”1. Defineupto4researchobjectives2. Teamassignment
<lunch>
1. ShortintroductionintothetoolFor each team:2. (Online)Surveydesignand implementation3. Piloting4. Wrap-up5. (Optional:Running survey with ESEMparticipants)
The“Plan”1. Defineupto4researchobjectives2. Teamassignment
<lunch>
1. ShortintroductionintothetoolFor each team:2. (Online)Surveydesignand implementation3. Piloting4. Wrap-up5. (Optional:Running survey with ESEMparticipants)
The“Plan”1. Defineupto4researchobjectives2. Teamassignment
<lunch>
1. ShortintroductionintothetoolFor each team:2. (Online)Surveydesignand implementation3. Piloting4. Wrap-up5. (Optional:Running survey with ESEMparticipants)
Wrap-Up
Eachgroup:Brieflyintroducethesurvey(researchobjective,targetpopulation,questionnaire)
à Askquestionsà Seewhatyoucanlearnfromotherdesigns
Alltogether• ShallwerunoneofthesurveysduringESEM?
15-20 minutes
Session IV
Q&A
Furtherreading…
• Groves, Fowler, Couper, Lepkowski, Singer and Torangeau, 2009. “Survey Methodology – 2nd edition” John Wiley and Sons
• Conradi R., Li J., Slyngstad O. P. N., Kampenes V. B., Bunse C., Morisio M., Torchiano M. “Reflections on conducting an international survey of CBSE in ICT industry” IEEE 4th International Symposium on Empirical Software Engineering November, 2005
• Linåker, J. Sulaman, S. M., de Mello, R. M. and Höst, M. 2015. Guidelines for Conducting Surveys in Software Engineering. TR 5366801, Lund University Publications. http://lup.lub.lu.se/record/5366801/file/5366839.pdf
• de Mello, R. M. and Travassos, G. H. 2016. Surveys in Software Engineering: Identifying Representative Samples. Proc. of 10th ACM/IEEE ESEM, Ciudad Real.
• de Mello, R. M. 2016. Conceptual Framework for Supporting the Identification of Representative Samples for Surveys in Software Engineering. Doctoral Thesis, COPPE/UFRJ, 138p. http://www.cos.ufrj.br/uploadfile/publicacao/2611.pdf