Applied Choice Analysis: A Primer

743 Pages·2005·5.563 MB·English

by  David A. Hensher, John M. Rose, William H. Greene

This page intentionally left blank
Applied Choice Analysis
A Primer
Almostwithoutexception,everythinghumanbeingsundertakeinvolvesachoice.
Inrecentyears,therehasbeenagrowinginterestinthedevelopmentandappli-
cation of quantitative statistical methods to study choices made by individuals
withthepurposeofgainingabetterunderstandingbothofhowchoicesaremade
and of forecasting future choice responses. In this primer, the authors provide
anunintimidatingintroductiontothemaintechniquesofchoiceanalysisandin-
cludedetailonthemessuchasdatacollectionandpreparation,modelestimation
andinterpretation,andthedesignofchoiceexperiments.Acompanionwebsite
to the book provides practice data sets and software to estimate the main dis-
cretechoicemodelssuchasmultinomiallogit,nestedlogit,andmixedlogit.This
primerwillbeaninvaluableresourcetostudentsaswellofimmensevaluetocon-
sultants/professionals,researchers,andanyoneelseinterestedinchoiceanalysis
andmodeling.
Companionwebsitewww.cambridge.org/0521605776
 . isDirectoroftheInstituteofTransportStudiesandProfes-
sorofManagementintheFacultyofEconomicsandBusinessattheUniversity
ofSydney
 . isaLecturerattheInstituteofTransportStudiesattheUniversity
ofSydney
 .  isProfessorofEconomicsandEntertainmentandMedia
FacultyFellowintheDepartmentofEconomicsattheSternSchoolofBusiness,
NewYorkUniversity

Applied Choice
Analysis
A Primer
David A. Hensher
TheUniversityofSydney
John M. Rose
TheUniversityofSydney
William H. Greene
NewYorkUniversity
  
Cambridge, NewYork, Melbourne,Madrid, CapeTown, Singapore, São Paulo
Cambridge University Press
TheEdinburghBuilding,Cambridge,UK
Published in theUnited Statesof America by Cambridge University Press, New York
www.cambridge.org
Information on this title: www.cambridg e.org /9780521844260
©DavidA.Hensher,JohnM.Rose,WilliamH.Greene2005
Thisbookisincopyright.Subjecttostatutoryexceptionandtotheprovisionof
relevantcollectivelicensingagreements,noreproductionofanypartmaytakeplace
withoutthewrittenpermissionofCambridgeUniversityPress.
Firstpublishedinprintformat 2005
- ---- eBook (NetLibrary)
- --- eBook (NetLibrary)
- ---- hardback
- --- hardback
- ---- paperback
- --- paperback
CambridgeUniversityPresshasnoresponsibilityforthepersistenceoraccuracyof
sforexternalorthird-partyinternetwebsitesreferredtointhisbook,anddoesnot
guaranteethatanycontentonsuchwebsitesis,orwillremain,accurateorappropriate.
Contents
Listoffigures pagexiii
Listoftables xviii
Preface xxiii
PartI Basictopics
1 Inthebeginning 3
2 Basicnotionsofstatistics 8
2.1 Introduction 8
2.2 Data 8
2.2.1 Theimportanceofunderstandingdata 10
2.3 Anoteonmathematicalnotation 10
2.3.1 Summation 11
2.3.2 Product 12
2.4 Probability 12
2.4.1 Relativefrequencies 13
2.4.2 Definingrandomvariables 14
2.4.3 Probabilitydistributionfunctions 14
2.4.4 Cumulativedistributionfunctions 16
2.4.5 Multivariateprobabilitydensityfunctions 17
2.4.6 Themultivariateprobabilityfunction 18
2.4.7 Marginalprobabilitydensityfunctions 21
2.4.8 Conditionalprobabilitydensityfunctions 21
2.4.9 Definingstatisticalindependence 23
2.5 Propertiesofrandomvariables 23
2.5.1 Expectedvalue 25
2.5.1.1 Propertiesofexpectedvalues 26
2.5.2 Variance 28
2.5.2.1 Propertiesofvariance 28
v
vi Contents
2.5.3 Covariance 30
2.5.3.1 Propertiesofcovariance 31
2.5.4 Thevariance–covariancematrix 32
2.5.5 Correlation 33
2.5.5.1 Propertiesofthecorrelationcoefficient 34
2.5.6 Correlationandvariances 36
2.6 Samplepopulationstatistics 36
2.6.1 Thesamplemean 36
2.6.2 Thesamplevariance 38
2.6.3 Thesamplecovariance 38
2.6.4 Thesamplecorrelationcoefficient 39
2.7 Samplingerrorandsamplingdistributions 39
2.8 Hypothesistesting 41
2.8.1 Definingthenullandalternativehypotheses 42
2.8.2 Selectingthetest-statistic 44
2.8.3 Significanceofthetestandalpha 45
2.8.4 Performingthetest 51
2.8.5 Examplehypothesistest:theonesamplet-test 51
2.9 Matrixalgebra 52
2.9.1 Transposition 53
2.9.2 Matrixadditionandsubtraction 53
2.9.3 Matrixmultiplicationbyascalar 54
2.9.4 Matrixmultiplication 54
2.9.5 Determinantsofmatrices 55
2.9.6 Theidentitymatrix 56
2.9.7 Theinverseofamatrix 57
2.9.8 Linearandquadraticforms 58
2.9.9 Positivedefiniteandnegativedefinitematrices 59
2.10 Conclusion 59
Appendix2A Measuresofcorrelationorsimilarity 59
3 Choosing 62
3.1 Introduction 62
3.2 Individualshavepreferences,andtheycount 63
3.3 Usingknowledgeofpreferencesandconstraintsinchoiceanalysis 71
3.4 Settingupabehavioralchoicerule 74
3.5 Derivingabasicchoicemodel 82
3.6 Concludingoverview 86
4 Paradigmsofchoicedata 88
4.1 Introduction 88
4.2 Dataconsistentwithchoice 89
4.3 Revealedpreferencedata 92
4.3.1 Choice-basedsampling 95
Contents vii
4.4 Statedpreference(orstatedchoice)data 96
4.5 Furthercomparisons 97
4.6 WhynotusebothRPandSPdata? 98
4.7 Socio-demographiccharacteristicdata 98
5 Processesinsettingupstatedchoiceexperiments 100
5.1 Introduction 100
5.2 Whatisanexperimentaldesign? 100
5.2.1 Stage1:Problemdefinitionrefinement 103
5.2.2 Stage2:Stimulirefinement 104
5.2.2.1 Refiningthelistofalternatives 104
5.2.2.2 Refiningthelistofattributesandattributelevels 105
5.2.3 Stage3:Experimentaldesignconsiderations 109
5.2.3.1 Labeledversusunlabeledexperiments 112
5.2.3.2 Reducingthenumberoflevels 114
5.2.3.3 Reducingthesizeofexperimentaldesigns 115
5.2.3.4 Dummyandeffectscoding 119
5.2.3.5 Calculatingthedegreesoffreedomrequired 122
5.2.3.6 Blockingthedesign 126
5.2.4 Stage4:Generatingexperimentaldesigns 127
5.2.4.1 Assigninganattributeasablockingvariable 130
5.2.5 Stage5:Allocatingattributestodesigncolumns 131
5.3 Anoteonunlabeledexperimentaldesigns 150
5.4 Optimaldesigns 152
Appendix5A Designingnestedattributes 154
Appendix5B Assignmentofquantitativeattribute-levellabels 156
6 Choicesindatacollection 161
6.1 Introduction 161
6.2 Generalsurveyinstrumentconstruction 161
6.3 Questionnairesforchoicedata 166
6.3.1 Stage6:Generationofchoicesets 166
6.3.2 Stage7:Randomizingchoicesets 170
6.3.3 Stage8:Surveyconstruction 172
6.3.3.1 Choicecontext 173
6.3.3.2 Useanexample 174
6.3.3.3 Independenceofchoicesets 174
6.3.3.4 Morethanonechoice 175
6.3.3.5 Theno-choiceordelay-choicealternative 176
6.4 Revealedpreferencesinquestionnaires 177
6.5 StudiesinvolvingbothRPandSPdata 177
6.6 UsingRPdatainSPexperiments:the“currentalternative” 178
6.7 Samplingforchoicedata:thetheory 184
6.7.1 Simplerandomsamples 185
viii Contents
6.7.2 Stratifiedrandomsampling 190
6.7.3 Conclusiontothetheoryofcalculatingsamplesizes 192
6.8 Samplingforchoicedata:thereality 193
7 NLOGITforappliedchoiceanalysis:aprimer 197
7.1 Introduction 197
7.2 Aboutthesoftware 197
7.2.1 AboutNLOGIT 197
7.2.2 AboutNLOGIT/ACA 198
7.2.3 InstallingNLOGIT/ACA 198
7.3 StartingNLOGIT/ACAandexitingafterasession 198
7.3.1 Startingtheprogram 198
7.3.2 Inputtingthedata 198
7.3.3 Readingdata 200
7.3.4 Theprojectfile 200
7.3.5 Leavingyoursession 201
7.4 UsingNLOGIT 201
7.5 HowtogetNLOGITtodowhatyouwant 202
7.5.1 UsingtheTextEditor 202
7.5.2 Commandformat 204
7.5.3 Commands 205
7.5.4 UsingtheProjectFileBox 206
7.6 Usefulhintsandtips 206
7.6.1 LimitationsinNLOGIT(andNLOGIT/ACA) 207
7.7 NLOGITsoftware 207
7.7.1 Support 208
7.7.2 Theprograminstalledonyourcomputer 208
7.7.3 UsingNLOGIT/ACAintheremainderofthebook 208
Appendix7A Diagnosticanderrormessages 208
8 Handlingchoicedata 218
8.1 Introduction 218
8.2 Thebasicdatasetup 219
8.2.1 Enteringmultipledatasets:stackingandmelding 222
8.2.2 Handlingdataonthenon-chosenalternativeinRPdata 222
8.2.3 Combiningsourcesofdata 224
8.2.4 Weightingonanexogenousvariable 226
8.2.5 Handlingrejection:the“nooption” 227
8.3 EnteringdataintoNLOGIT 230
8.3.1 EnteringdatadirectlyintoNLOGIT 230
8.3.2 ImportingdataintoNLOGIT 232
8.3.2.1 TheText/DocumentEditor 232
8.3.3 ReadingdataintoNLOGIT 232
8.3.4 WritingdataintoNLOGIT 235
8.3.5 Savingdatasets 235