Properties of Preference Questions: Utility Balance, Choice Balance, Configurators, and M-Efficiency

Save this PDF as:
 WORD  PNG  TXT  JPG

Size: px
Start display at page:

Download "Properties of Preference Questions: Utility Balance, Choice Balance, Configurators, and M-Efficiency"

Transcription

1 Ppeties f Pefeence Questins: Utility Balance, Chice Balance, Cnfiguats, and M-Efficiency by Jhn R. Hause and Olivie Tubia August, 3 Jhn R. Hause is the iin Pfess f Maketing, Slan Schl f Management, Massachusetts Institute f Technlgy, E56-34, 38 Memial Dive, Cambidge, MA 4, (67) 53-99, fax (67) , Olivie Tubia is a gaduate student at the Maketing Gup, Massachusetts Institute f Technlgy, E56-345, 38 Memial Dive, Cambidge, MA 4, (67) , fax (67) , This eseach was suppted by the MIT Slan Schl f Management and MIT s Cente f Innvatin in Pduct Develpment. This pape may be dwnladed fm the Vitual Custme Initiative s website, mitslan.mit.edu/vc. This website als cntains demnstatins f many f the pefeence measuement questins discussed in this pape, including cnfiguats. Special thanks t Eic Badlw, Theds Evgeniu, and Duncan Simeste f insightful cmments and Ely Dahan wh was instumental in develping the cnfiguat with which we cllected the data f the stylized fact n wam-up questins. Lim Weisbeg pvided the web design f the cnfiguat and the examples f paied-cmpaisn and chice-based questins. The auths ae listed in alphabetical de. Cntibutins wee equal and synegistic.

2 Ppeties f Pefeence Questins: Utility Balance, Chice Balance, Cnfiguats, and M-Efficiency Abstact This pape uses stylized mdels t exple efficient pefeence questins. The fmal analyses pvide insights n citeia that ae cmmn in widely-used questin-design algithms. We demnstate that the citein f metic utility balance leads t inefficient questins, abslute biases, and elative biases amng patwths. On the the hand, chice balance f statedchice questins (the analgy f utility balance f metic questins) can lead t efficiency unde sme cnditins, especially f discete featues. When at least ne fcal featue is cntinuus, chices ae imbalanced at ptimal efficiency. This chice imbalance declines as espnses becme me accuate and nn-fcal featues incease. The analysis f utility- and chice-balance pvide insights n the elative success f metic- and chice-based adaptive plyhedal methds and suggest futue impvements. We then use the fmal mdels t exple a new fm f data cllectin that has becme ppula in e-cmmece and web-based maket eseach cnfiguats. We study the stylized fact that questins nmally used as wam-up questins in cnjint analysis ae supising accuate in pedicting cnsume behavi that is elevant t e-cmmece applicatins. F these applicatins, new data suggest that wam-up questins ae, pehaps, me useful manageially than paied-cmpaisn tadeff questins. We examine and genealize this insight t demnstate the advantages f matching tadeff, cnfiguat, and ange questins t the elevant manageial decisins. This futhe genealizes t M-efficiency (manageial efficiency), an altenative citeia f questin design algithms. Designs which minimize D-es and A-es may nt minimize M D - and M A -es which measue pecisin elevant t manageial decisins. We pvide examples in which simple mdificatins t questin design incease M-efficiency. eywds: Cnjint analysis, efficient questin design, adaptive questin design, Intenet maket eseach, e-cmmece, pduct develpment.

3 Utility Balance, Chice Balance, Cnfiguats, and M-Efficiency. Mtivatin Pefeence measuement is cucial t manageial decisins in maketing. Methds such as cnjint analysis, vice-f-the-custme methds, and, me ecently, cnfiguats, ae used widely f pduct develpment, advetising, and stategic maketing. In espnse t this manageial need, maketing scientists have develped many sphisticated and successful algithms t select questins efficiently. F example, Figue illustates tw ppula fmats that have been studied widely. Figue Ppula Fmats f Cnjint Analysis (a) Metic paied-cmpaisn fmat (laptp bags) (b) Stated-chice fmat (Executive Educatin) F the metic-pais fmat (Figue a) questins can be selected with efficient designs, with Adaptive Cnjint Analysis (ACA), with plyhedal methds (uhfield, Tbias, and Gaatt 994; Sawtth ; Tubia, Simeste, Hause, Dahan 3). F the stated-chice fmat (Figue b), ecent advances adapt questins based n petests, manageial beliefs, pi questins and d s eithe by selecting an aggegate design that is eplicated f all subsequent espndents by adapting questin design f each espndent (Aa and Hube ; Hube and Zweina 996; Jhnsn, Hube and Bacn 3; anninen ; Sand and Wedel,, Tubia, Hause, and Simeste 3). These questin-design methds ae used with a vaiety f estimatin methds (egessin, Hieachical Bayes estimatin, mixed lgit mdels, analytic cente appximatins, and hybid methds) and have pven t pvide me efficient and me accuate data with which t estimate patwths.

4 Utility Balance, Chice Balance, Cnfiguats, and M-Efficiency Figue Example Cnfiguats (a) Cnfiguat fm Dell.cm (b) Innvatin Tlkit f Industial Flavings (c) Design Palette f Pickup Tucks (d) Use Design f Laptp Cmpute Bags Recently, as pefeence measuement has mved t web-based questinnaies, e- cmmece fims and academic eseaches have expeimented with a new fmat in which espndents ( e-cmmece custmes) select featues t cnfigue a pefeed pduct. The bestknwn e-cmmece example is Dell.cm (Figue a); hweve, the fmat is becming ppula n many websites, especially autmtive websites (e.g., kbb.cm s aut chice advis, vehix.cm, autweb.cm). F maketing eseach, cnfiguats ae knwn vaiusly as innvatin tlkits, design palettes, and use design fmats. They have been instumental in the design f flavings f Intenatinal Flav and Fagance, Inc (Figue b), new tuck platfms f Geneal Mts (Figue c), and new messenge bags f Timbuk (Figue d). (Figues fm Dahan and Hause ; Thmke and vn Hippel, Uban and Hause 3; vn Hippel ). Respndents find cnfiguats easy t use and a natual means t expess pefeences. Hweve, because cnfiguats fcus n the espndents pefeed pduct, they have nly

5 Utility Balance, Chice Balance, Cnfiguats, and M-Efficiency been used t estimate patwth estimates when espndents ae asked t epeat the tasks with vaying stimuli (e.g., Liechty, Ramaswamy and Chen ). In this pape, we examine questin design f all thee fms f pefeence measuement: metic pais, stated chices, and cnfiguats. We use fmal mdeling methds t examine key ppeties that have been used by the algithms ppsed t date. With these mdels we seek t undestand basic ppeties in de t help explain why sme algithms pefm well and thes d nt. These mdels als pvide insight n the le f cnfiguats in pefeence measuement and indicate the stengths and weaknesses f this new fm f data cllectin elative t me-standad methds. The mdels themselves ae dawn fm empiical expeience and epesent, in a stylized manne, the basic ppeties f questin design. The fmal mdels als pvide insight n the elatinship f questin design t manageial use suggesting, f example, that sme measuement fmats ae best f develping a single pduct, sme f pduct-line decisins, and sme f e-cmmece decisins. These insights lead t evised manageial citeia (M A -e and M D -e). The pape is stuctued as fllws: We examine fist the cncept f utility balance as applied t the metic paiedcmpaisn fmat and demnstate that this citein leads t bth bias and inefficiency f that fmat. This analysis als lends insight n ecent cmpaisns between ACA and plyhedal methds (Tubia, et. al. 3). We then examine a elated cncept f the stated-chice fmat. The liteatue ften labels this cncept utility balance, but, f the pupse f this pape, we label it chice balance t indicate its elatinship t stated-chice data. Ou analyses indicate that, unlike utility-balance f metic data, chice balance f stated-chice data can lead t impved efficiency unde sme cnditins. In the cases, ptimal efficiency equies unbalanced chices. Ou analyses bth eflect insights fm ecent algithmic papes and pvide new insights n hw chice balance changes as a functin f the chaacteistics f the pblem. We then examine data that suggests that cnfiguat questins pvide me useful infmatin f e-cmmece decisins than tadeff questins (Figue ). We exple this bsevatin with a stylized mdel t suggest hw questin fmats ae best matched t manageial decisins. These insights genealize t M-efficiency. 3

6 Utility Balance, Chice Balance, Cnfiguats, and M-Efficiency We define u scpe, in pat, by what we d nt addess. Althugh we daw expeience fm and seek t pvide insight f new algithms, algithmic design is nt the fcus f this pape. N d we wish t ente the debate n the elative meits f metic-ating stated-chice data. F nw we leave that debate t the auths such as Eld, Luviee and Davey (99) Me (3). We study bth data fmats (and cnfiguats) and use fmal mdels t gain insights n all thee fms f pefeence questins. We d nt addess hieachical mdels f espndent hetegeneity, but athe fcus n each espndent individually. We ecgnize the pactical imptance f hieachical mdels, but abstact fm such estimatin t fcus n the basic ppeties. The ecent emegence f individual-specific questin designs and/ diffeentiated designs suggest that u insights shuld extend t such analyses (Jhnsn, Hube, and Bacn 3; Sand and Wedel 3; Tubia, Hause, and Simeste 3). But that emains futue eseach.. Efficiency in Questin Design Befe we begin u analyses, we eview biefly the cncepts f efficient questin design. The cncept f questin efficiency has evlved t epesent the accuacy with which patwths ae estimated. Efficiency fcuses n the standad es f the estimates. Let p be the vect f patwths t be estimated and pˆ the vect f estimated patwths. Then, if pˆ is (appximately) nmally distibuted with vaiance Σ, the cnfidence egin f the estimates is an ellipsid defined by ( p p)' Σ ˆ ( p pˆ), Geene (993, p. 9). F mst estimatin methds, Σ depends upn the questins that the espndent is asked and, hence, an efficient set f questins minimizes the cnfidence ellipsid. This is implemented as minimizing a nm f the matix, Σ. A-es ae based n the tace f Σ, D-es n the deteminant f Σ, and G-es n the maximum diagnal element. Because the deteminant is ften the easiest t ptimize, mst empiical algithms wk with D-efficiency (uhfeld, Tbias, and Gaatt 994, p. 547). G- efficiency is aely used. Because the deteminant f a matix is the pduct f its eigenvalues and the tace is the sum f its eigenvalues, A-es and D-es ae ften highly celated (uhfeld, Tbias, and Gaatt 994, p. 547, Aa and Hube, p. 74). In this pape we use bth, chsing the ci- In pactice, Σ is unknwn and is eplaced with its estimated value. F simplicity we assume that Σ is knwn, which shuld change neithe the insights n the esults f u analyses. 4

7 Utility Balance, Chice Balance, Cnfiguats, and M-Efficiency tein which leads t the mst tanspaent insight. In sme cases the esults ae easie t explain if we wk diectly with Σ ecgnizing det(σ) = [det( Σ )]. 3. Utility Balance f the Metic Paied-Cmpaisn Fmat The metic pais fmat (Figue a) is widely used f cmmecial cnjint analysis (Geen, iege, Wind, p. S66; Te Hfstede, im and Wedel, p. 59). Althugh the fmat has been used academically and cmmecially f ve 5 yeas, its mst cmmn applicatin is in the adaptive cmpnent f ACA (Geen, iege and Agawal 99, Hause and Shugan 98, Jhnsn 99). 3 Metic utility balance is at the heat f ACA. With the gal that a difficult chice pvides bette infmatin f futhe efining utility estimates, ACA pesents t the espndent pais f cncepts that ae as nealy equal as pssible in estimated utility (Ome 999, p. ; Sawtth Sftwae,, p., espectively). This citein has appeal. F example, in the study f espnse latencies, Haaije, amakua, and Wedel (, p. 38) suggest that espndents take me time n chice sets that ae balanced in utility and, theefe, make less epne chices. Geen, iege and Agawal (99, p. 6) quting a study by Hube and Hansen (986) ascibe bette pedictive ability t geate espndent inteest in these difficult-t-judge pais. Finally, Shugan (98) pvides a they t suggest that the cst f thinking is invesely pptinal t the squae f the utility diffeence. Thee ae clea advantages in tems f espndent inteest, attentin, and accuacy f metic utility-balanced questins. Hweve, ecent simulatin and empiical evidence suggests that altenative citeia pvide questin designs that lead t me accuate patwth estimates (Tubia, et. al. 3). Futheme, adaptatin implies endgeneity in the design matix, which, in geneal, leads t bias. Thus, we exple whethe nt the citein f metic utility balance has advese chaacteistics, such as systematic biases, that ffset the advantages in espndent attentin. We begin with a stylized mdel. The tace f a matix des nt necessaily equal the invese f the tace f the invese matix, thus an empiical algithm that maximizes the tace f Σ - des nt necessaily minimize the tace f Σ. Hweve, the eigenvalues ae elated, especially when the eigenvalues f Σ ae simila in magnitude, as they ae in empiical situatins. 3 F example, D. Lynd Bacn, AMA s Vice Pesident f Maket Reseach and Executive Vice Pesident f NFO Wldgup cnfims that metic pais emain ne f the mst ppula fms f cnjint analysis. Sawtth Sftwae claims that ACA is the mst ppula fm f cnjint analysis in the wld, likely has the lagest installed base, and, in, accunted f 37% f thei sales. Pivate cmmunicatins and Maketing News, 4//, p.. 5

8 Utility Balance, Chice Balance, Cnfiguats, and M-Efficiency A Stylized Mdel F simplicity we cnside pducts with tw featues, each f which is specified at tw levels. We assume futhe that the featues satisfy pefeential independence such that we can wite the utility f a pfile as an additive sum f the patwths f the featues f the pfile (eeney and Raiffa 976, p. -5; antz, et. al. 97). Withut lss f geneality, we scale the lw level f each featue t ze. Let p be the patwth f the high level f featue and let p be the patwth f the high level f featue. F these assumptins thee ae fu pssible pfiles which we wite as {, }, {, }, {, }, and {, } t indicate the levels f featues and, espectively. With this ntatin we have the tue utilities given by: u(, ) = u(, ) = p u(, ) = p u(, ) = p p F metic questins in which the espndent pvides an inteval-scaled espnse, we mdel espnse e as an additive, ze-mean andm vaiable, e. F example, if the espndent is asked t cmpae {, } t {, }, the answe is given by a p p e. We dente = the pbability distibutin f e with f (e). We dente the estimates f the patwths, estimated fm espnses t the questins, by ˆp and ˆp. Adaptive (Metic) Utility Balance Withut lss f geneality, cnside the case whee p > p > p such that the ff- > diagnal questin ({,} vs. {,}) is the mst utility-balanced questin. We assume that this dinal elatinship is based n pi questins. We label the e assciated with the mst utility-balanced questin as eub and label es assciated with subsequent questins as eithe e e. We nw cnside an algithm which adapts questins based n utility balance. Adaptive metic utility balance implies the fllwing sequence. Fist questin: p p = p p e ˆ ˆ ub p ˆ = p e e ub < p p p ˆ = p e e ub p p Secnd questin: if (Case ) Suppse that e ub p p, then: E [ pˆ = p ] = p E[ e ] if (Case ) E[ pˆ ˆ p ] = E[ p ] p p E[ eub eub p p] = p E[ eub eub p p] > 6

9 Utility Balance, Chice Balance, Cnfiguats, and M-Efficiency Suppse that e ub < p p, then: E [ pˆ = p ] = p E[ e ] E[ pˆ ˆ p ] = E[ p] p p E[ eub eub < p p] = p E[ eub eub < p p] > Thus, unde bth Case and Case, ne f the patwths is biased upwads f any ze-mean distibutin that has nn-ze measue belw p - p. The intuitin undelying this pf is that the secnd questin in the adaptive algithm depends diectly n the andm e a diect vesin f endgeneity bias (Judge, et. al. 985, p. 57). We demnstate hw t genealize this esult with a pseud-egessin fmat f ˆ p = ( X ' X ) ( X ' a) = p ( X ' X ) X ' e, whee a and e ae the answe and e vects and X is the questin matix. Then the secnd w f X depends n the magnitude f eub elative t p - p. Specifically, the fist w f X is [,-] and its secnd w can be witten as [b, -b] whee b = if es ae small and b = if es ae lage. The bias, B, is then given by the fllwing equatin whee k =. ( ) ( ) b e ub ek b eub () E[ B ] = E[( X ' X ) X ' e] = E( ) = E( ) beub ek beub Equatin illustates the phenmenn clealy: b = if e ub is small (me negative) and b = if e ub is lage (me negative), hence eithe ˆp ˆp is biased upwad. These aguments genealize eadily t me than tw featues and me than tw questins. The basic insight is that the es fm ealie questins detemine the late questins (X) in a systematic way the essence f endgeneity bias. Relative Bias Bias alne des nt necessaily impact manageial decisins because patwths ae unique nly t a psitive linea tansfmatin. Thus, we need t establish that the bias is als elative, i.e., the bias depends n the magnitude f the tue patwth. Because the bias is p- ptinal t e ub we need t shw that p E[ eub eub p p] < t demnstate that bias is geate f smalle (tue) patwths than it is f lage (tue) patwths. We d s by diect diffeentiatin in the Appendix. Thus, we can state the fllwing ppsitin. 7

10 Utility Balance, Chice Balance, Cnfiguats, and M-Efficiency Ppsitin. F the stylized mdel, n aveage, metic utility balance biases patwths upwad and des s diffeentially depending upn the tue values f the patwths. When we expand the adaptive questins t a me cmplex wld, in which thee ae me than tw featues, the intuitin still hlds; hweve, the magnitude f the bias depends upn the specific manne by which metic utility balance is achieved and n the accuacy f the espndent s answes. As an illustatin, we built a 6-questin simulat f a tw-patwth pblem in which, afte the fist questin, all subsequent questins wee chsen t achieve metic utility balance based n patwths estimated fm pi questins. Based n 6, egessins each, biases anged fm % f elatively lage tue patwths t 3% f elatively small patwths. 4 T examine the pactical significance we cite data fm Tubia, et. al. (3) wh simulate ACA questin design f a -patwth pblem. In thei data, biases wee appximately 6.6% f the magnitudes f the patwths at eight questins when aveaged acss the cnditins in thei expeimental design. The biases ae significant at the.5 level. The Effect f Metic Utility Balance n Efficiency Althugh metic utility balance leads t bias, it might be the case that adaptatin leads t geate efficiency. F D-es, minimizing the deteminant f Σ is the same as maximizing the deteminant f Σ. F metic data, Σ = X ' X (f apppiately centeed X). Fm this expessin we see tw things immediately. Fist, as anninen (, p. 7) and uhfeld, et. al. (994, p. 547) nte, f a given set f levels, efficiency tends t push featues t thei exteme values. (These statements assume that the maximum levels ae selected t be as lage as is easnable f the empiical pblem. Levels which ae t exteme isk additinal espndent e nt explicitly acknwledged by the statistical mdel used in questin design. See Delquie 3). Secnd, pefect metic utility balance impses a linea cnstaint n the clumns f X because X p =. This cnstaint induces X X t be singula. When X X is singula, the deteminant will be ze and D-es incease withut bund. Impefect metic utility balance leads t 4 The simulat is available fm the website listed in the acknwledgements sectin f this pape. The simulat is designed t demnstate the phenmenn; we ely n Tubia, et. al. (3) f an estimate f the empiical magnitude. The simulat chses x unifmly n [,] and x based n utility balance. The abslute values f the espnse es ae appximately half thse f the dependent measue. The magnitude f the bias changes if the espnse es change, but the diectin f the bias emains. 8

11 Utility Balance, Chice Balance, Cnfiguats, and M-Efficiency det(x X) and extemely lage D-es. A-es behave similaly. Thus, geate metic utility balance tends t make questins inefficient. 5 Insights n the Relative Pefmance f Metic Plyhedal Methds The analysis f metic utility balance pvides insights n the elative pefmance f metic plyhedal methds. Metic plyhedal methds fcus n asking questins t educe the feasible set f patwths as apidly as pssible by slving an eigenvalue pblem elated t Snnevend ellipsids. These ellipsids fcus questins elative t the axis abut which thee is the mst uncetainty. Plyhedal methds d nt attempt t impse metic utility balance. In algithms t date, plyhedal methds ae used f (at mst) the fist n questins whee n equals the numbe f patwths. Hweve, metic plyhedal methds als impse the cnstaint that the ws f X be thgnal (Tubia, et. al. 3, Equatin A9). Thus, afte n questins, X will be squae, nn-singula, and thgnal ( XX ' I implies X ' X I ). 6 Subject t scaling, this thgnality elatinship minimizes D-e (and A-e). Thus, while the adaptatin inheent in metic plyhedal methds leads t endgenus questin design, the thgnality cnstaint appeas t vecme the tendency f endgeneity t incease es. Tubia, et. al. (3) ept at mst a % bias f metic plyhedal questin selectin, significantly less than bseved f ACA questin selectin. Summay Metic utility balance may etain espndent inteest and encuage espndents t think had abut tadeffs, hweve, these benefits cme at a pice. Metic utility balance leads t bias, elative bias, and lweed efficiency. On the the hand, the inheent thgnality citein (athe than utility balance) in metic plyhedal methds enables metic plyhedal questin design t achieve the benefits f adaptatin with significantly less bias and, appaently, withut educed efficiency. 4. Chice Balance f the Stated-Chice Fmat In the stated-chice fmat (Figue b), tw pfiles that have the same utility will be equally likely t be chsen, thus chice balance is elated t utility balance. Hweve, because 5 The simulat cited in the acknwledgements sectin f this pape als cntains an example in which metic utility balance can be impsed. The me clsely utilities ae balanced, the wse the fit. 6 Othgnal ws assume that XX =αi whee α is a pptinality cnstant. Because X is nn-singula, X and X ae invetible, thus X XX X - = αx X - X X = αi. 9

12 Utility Balance, Chice Balance, Cnfiguats, and M-Efficiency the undelying statistical mdels ae fundamentally diffeent, the implicatins f chice balance f stated-chice questins ae diffeent fm the implicatins f metic utility f metic paiedcmpaisn questins. Chice balance is an imptant citein in many f the algithms used t custmize stated-chice questins. F example, Hube and Zweina (996) begin with thgnal, levelbalanced, minimal velap designs and demnstate that swapping and elabeling these designs f geate chice balance inceases D p -efficiency. Aa and Hube (, p. 75) test these algithms in a vaiety f dmains with hieachical Bayes methds cmmenting that the maximum D p -efficiency is achieved by sacificing thgnality f bette chice balance. 7 Ome and Hube (, p. 8) futhe advcate chice balance. Tubia, Hause, and Simeste (3) use chice-balance t incease accuacy in adaptive chice-based questins. Finally, the cncept f chice balance is gemane t the pinciple, used in the cmpute adaptive testing liteatue, that a test is mst effective when its items ae neithe t difficult n t easy, such that the pbabilities f cect espnses ae clse t.5 (Ld 97). T undestand bette why chice balance impves efficiency while metic utility balance degades efficiency, we ecgnize that, f lgit estimatin f patwths with stated-chice data, Σ depends upn the tue patwths and, implicitly, n espnse e. Specifically, McFadden (974) shws that: q J i i i () Σ = R ( xij xik Pik )' Pij ( xij J i= j= k= k= whee R is the effective numbe f eplicates; J i is the numbe f pfiles in chice set i; q is the numbe f chice sets; x ij is a w vect descibing the j th pfile in the i th chice set; and P ij is the pbability that the espndent chses pfile j fm the i th chice set. Equatin has a me tanspaent fm when all chice sets ae binay, in paticula: q (3) Σ = R ( xi xi )' Pi ( Pi )( xi xi ) = R whee d i i= i= J x ik P ik ) q d ' P ( P is the w vect f diffeences in the featues f the i th chice set. In Equatin 3 it is clea that chice balance (P i ½) tends t incease a nm f Σ i i i ) d i. Hweve, the chice pb- 7 D p -efficiency ecgnizes that Σ depends n the tue patwths. See Equatin.

13 Utility Balance, Chice Balance, Cnfiguats, and M-Efficiency abilities (P i ) ae a functin f the featue diffeences ( d i ), thus pue chice balance may nt maximize eithe the deteminant ( the tace) f Σ. Fist ecgnize that the tue chice pbabilities in the lgit mdel ae functins f the utilities, d p. If we hld the utilities cnstant, we als hld the chice pbabilities cnstant. i Subject t this linea utility cnstaint, we can incease efficiency withut bund by inceasing the abslute magnitudes f the d i s. This esult is simila t that f the metic-pais fmat; the expeimente shuld chse featue diffeences that ae as lage as easnable subject t the linea cnstaint impsed by d i p and subject t any es (utside the mdel) intduced by lage featue diffeences (Delquie 3). anninen () explits these ppeties effectively by selecting ne cntinuus featue (e.g. pice) as a fcal featue with which t manipulate d p, while teating all emaining featues as discete by setting them t thei maximum minimum levels. She epts substantial inceases in efficiency f bth binay and multinmial chice sets (p. 3). anninen uses numeical means t select the ptimal levels f the cntinuus fcal featue. We gain futhe insight with fmal analysis. In paticula, we examine the ptimality cnditins f the fcal featue, hlding the thes cnstant. F geate tanspaency we use binay questins and the tace f Σ q d ik P i i= k = (4) tace( Σ ) = i athe than the deteminant. 8 The tace is given by: If we assume, withut lss f geneality, that the fcal featue is the th featue, then the fistde cnditins f the fcal featue ae: ( P i P (5) i d i = p dik f all i k = whee d ik is the level f the k th featue diffeence in the i th binay questin. We fist ule ut the tivial slutin f d ik = P i ) f all k. Nt nly wuld this imply a chice between tw identical altenatives, but the secnd-de cnditins imply a minimum. A slightly me inteesting case ccus when pefect chice balance is aleady achieved by the - 8 F a fcal featue, the tace and the deteminant have elated maxima, especially if all but the nn-fcal featue ae set t thei minimum maximum values.

14 Utility Balance, Chice Balance, Cnfiguats, and M-Efficiency discete featues. We etun t this slutin late, hweve, in the fcal-featue case it ccus with ze pbability (i.e., n a set f ze measue). The me inteesting slutins f a cntinuus fcal featue equie that d i be nn-ze and, hence, that P P i i Pi. Thus, when di can vay cntinuusly, Equatin 5 implies that ptimal efficiency equies chice imbalance. Thee ae tw equivalent slutins t Equatin 5 cespnding t a ight-t-left switch f the tw altenatives. That is, the sign f d i must match the sign f. Thus, withut lss f geneality, we examine situatins wee di is psitive (see Appendix). We plt the maginal impact f d i n tace( Σ ) f illustative levels f the nn-fcal featue diffeence. 9 Fllwing Aa and Hube () and Tubia, Hause and Simeste (3), we use the magnitude f the patwths as a measue f espnse accuacy highe magnitudes indicate highe espnse accuacy. Figue 3 illustates that efficiency is maximized f nn-ze fcal featue diffeences. Figue 3 als illustates that the ptimal level f the fcal featue diffeence depends upn the level f the nn-fcal featues and upn the magnitude. i i Chice balance ( P P ) is a nn-linea functin f the featue diffeences. Nnetheless, we can identify the systematic impact f bth nn-fcal featues and espnse accuacy. In paticula, fmal analyses eveal that chice balance inceases with geate espnse accuacy and with geate levels f the nn-fcal featue diffeences. We fmalize these insights in tw ppsitins. The pfs ae in the Appendix. P i Ppsitin. As espnse accuacy inceases, chice balance inceases and the level diffeence in the fcal featue deceases. Ppsitin 3. As the levels f the nn-fcal featues incease, chice balance inceases. 9 F lw magnitudes, the patwths f bth featue ae set equal t. in these plts. F high magnitudes they ae dubled. The lw level f the nn-fcal featue (abslute value) is.5. The high level is 3..

15 Utility Balance, Chice Balance, Cnfiguats, and M-Efficiency Figue 3 Optimal Efficiency and Chice Imbalance Efficicency (tace) 4 3 Efficicency (tace) Featue diffeence Lw levels nn-fcal featues High levels nn-fcal featues Featue diffeence Lw magnitude High magnitude (a) Lw vs. high nn-fcal featue levels (b) Lw vs. high magnitude in lgit mdel Quantal Featues In many applicatins, all featues ae binay they ae eithe pesent absent. Examples include fu-wheel steeing n autmbiles, cell-phne hldes n laptp bags, and aut fcus n cameas. In these cases, the fcal-featue methd cannt be used f maximizing efficiency. Hweve, discete-seach algithms ae feasible t maximize a nm n. Aa and Hube () and Sand and Wedel () pvide tw such algithms. Sand and Wedel () extend these cncepts t a mixed lgit mdel. An inteesting (and cmmn) case ccus when the expeimente limits the numbe f featues that vay, pehaps due t cgnitive lad (Sawtth Sftwae 996, p. 7; Shugan 98). Σ F a given numbe f binay featues (±), k d ik is fixed. In this case, it is easy t shw that Equatin 4 is maximized when chices ae as balanced as feasible. This same phenmenn applies when the eseache equies that all featues vay. Insights n the Relative Pefmance f Chice-based Plyhedal Methds Like metic plyhedal methds, chice-based plyhedal methds ely upn Snnevend ellipsids. Hweve, f stated-chice data, espndents answes pvide inequality cnstaints athe than equality cnstaints. Rathe than selecting questins vects paallel t the lngest axis f the ellipsid, chice-based plyhedal methds use the ellipsid t identify taget pat- If thee ae a maximum numbe f featues that can be nn-ze athe than a fixed numbe, it is pssible, althugh less likely, that a maximum will might ccu f less than the maximum numbe f featues. This case can be checked numeically f any empiical pblem. Nnetheless, the qualitative insight hlds. 3

16 Utility Balance, Chice Balance, Cnfiguats, and M-Efficiency wths at the extemes f the feasible egin. The chice-based algithm then slves a utility maximizatin (knapsack) pblem t identify the featues f the pfiles. The knapsack pblem assues that the cnstaints g (appximately) thugh the analytic cente f the egin. This citein pvides appximate a pii chice balance. (By a pii we mean that, given the pi distibutin f feasible pints, the pedicted chices ae equal.) In this manne, the algithm achieves chice balance withut equiing utility balance thughut the egin. Utility balance nly hlds f the analytic cente f the egin a set f ze measue. The applicatins f chice-based plyhedal methds t date have fcused n discete featues implemented as binay featues. F discete featues, chice balance achieves high efficiency. F cntinuus featues, Ppsitin suggests algithmic impvements. Specifically, the knapsack pblem might be mdified t include tace( Σ ) in the bjective functin. Hweve, unlike a knapsack slutin, which is btained apidly, the evised math pgam may need t ely n new, ceative heuistics. Summay Chice balance affects the efficiency f stated-chice questins quite diffeently than metic utility balance affects metic paied-cmpaisn questins. This, despite the fact that chice balance is elated t dinal utility balance. If at least ne featue is cntinuus, then that featue can be used as a fcal featue and the mst efficient questins equie chice imbalance (and nn-ze levels f the fcal featue). Hweve, we shw fmally that me accuate espnses and highe nn-fcal featues lead t geate chice balance. If featues ae discete and the numbe which can vay is fixed, then chices shuld be as balanced as feasible. Finally, this analysis helps explain the elative pefmance f stated-chice plyhedal methds and aggegate custmizatin methds. 5. Cnfiguats In Sectins 3 and 4 we expled tw ppula fmats f pefeence questins suggesting that, f the metic paied-cmpaisn fmat, utility balance led t bias, elative bias and inefficiency while, f the chice-based fmat, chice balance culd be efficient f discete featues but nt necessaily f cntinuus fcal featues. We nw exple a elatively new fmat, cnfiguats, in which espndents fcus n the featues ne at a time and ae nt asked t cmpae pfiles in the taditinal sense. Neithe utility balance n chice balance apply t cnfiguats. Nnetheless, analysis f this new fmat leads t insights that genealize t all thee questin 4

17 Utility Balance, Chice Balance, Cnfiguats, and M-Efficiency fmats and helps us undestand which fmat is mst efficient in which situatins. In cnfiguats, espndents ae typically given a pice f each featue level and asked t indicate thei pefeences f that pice/featue cmbinatin (eview Figue ). Cnfiguat questins ae equivalent t asking the espndent whethe the patwths f the featues (p k s) ae geate than less than the stated pices. These data appea quite useful. F example, Liechty, Ramaswamy and Chen () use ten epeated cnfiguat measues pe espndent t estimate successfully hieachical Bayes pbit mdels. Stylized Fact: Wam-up Questins d Supisingly Well in Pedicting Featue Chice Figue 4 pesents an empiical example. In data cllected by Tubia, et. al. (3), espndents wee shwn five laptp bags wth appximately $ and allwed t chse amng thse bags. Vaius cnjint methds based n metic paied-cmpaisn questins cllected pi t this task pedicted these pfile chices quite well. Figue 4 Wam-up Questins ae Effective Pedicts f Cnfiguat Tasks Pecent f Featues Pedicted Cectly With Initial Questins Withut Initial Questins Pedicted Pecentile f Chsen Pfile With Initial Questins Withut Initial Questins Numbe f Cnjint Questins Numbe f Cnjint Questins (a) Ability t pedict featue chice (b) Ability t pedict pefeed pfile Befe cmpleting the metic paied-cmpaisn task, in de t familiaize espndents with the featues, espndents answeed wam-up questins. They wee asked t indicate the pefeed level f each featue. Such wam-up questins ae cmmn. We e-analyzed Tubia, et. al. s data in tw ways. One cnjint estimatin methd did nt use the wam-up questins. The the cnjint estimatin methd included mntnic cnstaints based n the wam-up questins. One mnth afte the initial cnjint-analysis questinnaie, the espndents wee In the methd labeled with initial questins, the wam-up questins whee used t detemine which f tw featue levels wee pefeed (e.g., ed black). In the methd labeled withut initial questins, such deing data 5

18 Utility Balance, Chice Balance, Cnfiguats, and M-Efficiency ecntacted and given a chance t custmize thei chsen bag with a cnfiguat. Figue 4 plts the ability f the cnjint analyses t pedict cnfiguat chice as a functin f the numbe f metic paied-cmpaisn questins that wee answeed. F bth pedictive citeia, the ability t pedict cnfiguat chice is supising flat when initial wam-up questins ae included. (Pedictins ae less flat f fecasts f the chice f five laptp bags, Tubia, et. al. 3, Table 6.) F cnfiguat pedictins, the nine wamup questins appea t pvide me usable infmatin than the sixteen paied-cmpaisn questins. This empiical bsevatin is elated t bsevatins by Evgeniu, Bussis, and Zachaia (3) whse simulatins suggest the imptance f dinal cnstaints. While, at fist, Figue 4 appeas t be cunte-intuitive, it becmes me intuitive when we ealize that the wam-up questins ae simila t cnfiguat chice and the paied-cmpaisn questins ae simila t the chice amng five (Paet) pfiles. We fmalize this intuitin with u stylized mdel. Stylized Mdel t Explain the Efficiency f Wam-up Questins F tanspaency assume metic wam-up questins. This enables us t cmpae the wam-up questins t paied-cmpaisn questins withut cnfunding a metic-vs.-dinal chaacteizatin. We cnside the same thee questins used peviusly, ewiting e as e f mnemnic expsitin f paied-cmpaisn tadeff questins. Withut lss f geneality we incpate featue pices int the definitin f each featue. We assume p nt assume p > p as we did befe. The thee questins ae: (Q) p ˆ = p e (Q) p ˆ = p e (Q3) p p = p p e ˆ ˆ t > p > ub, but need We nw cnside stylized estimates btained fm tw questins. In u stylized wld f tw featues at tw levels, Q and Q can be eithe metic wam-up questins metic paied-cmpaisn questins. Q3 can nly be a metic paied-cmpaisn questin. Assuming all questins ae unbiased, we examine which cmbinatin f questins pvides the geatest pecisin (efficiency): Q Q, QQ3, QQ3. wee nt available t the estimatin algithm. In this paticula plt, f cnsistency, we use the analytic-cente methd f Tubia, et. al., hweve, the basic insights ae nt limited t this estimatin methd. t 6

19 Utility Balance, Chice Balance, Cnfiguats, and M-Efficiency Cnside featue. Because featue pice is incpated in the featue, if p k, then featue chice is cectly pedicted if pˆ k is nn-negative. (We use simila easning f featue.) Simple calculatins (see Appendix) eveal that the mst accuate cmbinatin f tw questins is QQ if and nly if Pb[ e p ] > Pb[ e e p ]. Since p is a negative k k k t k k numbe, these cnditins hld f mst easnable density functins. F example, in the Appendix we demnstate that, if the es ae independent and identically distibuted nmal andm vaiables, then this cnditin hlds. Ppsitin 4. F ze-mean nmally distibuted (i.i.d.) es, the wam-up questins (QQ) pvide me accuate estimates f featue chice than d pais f questins that include a metic paied-cmpaisn questin (Q3). Intuitively, metic wam-up questins pase the es me effectively than tadeff questins when the fcus is n featue chice athe than hlistic pducts. The questins fcus pecisin whee it is needed f the manageial decisins. We pvide a me fmal genealizatin f these cncepts in the next sectin. The intuitin des nt depend upn the metic ppeties f the wam-up questins and applies equally well t dinal questins. We invite the eade t extend Ppsitin 4 t dinal questins using the M-efficiency famewk that is expled in the next sectin. 6. Genealizatins f Questin Fcus: M-Efficiency The intuitin dawn fm Ppsitin 4 suggests that it is best t match pefeence questins caefully t the manageial task. F example, metic wam-up questins ae simila t cnfiguat questins and may be the mst efficient f pedicting featue chice. On the the hand, tadeff questins ae simila t the chice f a single pduct fm a Paet set. Estimating patwths fm tadeff questins may be best if the manageial gal is t pedict cnsume chice when a single new pduct is launched int an existing maket. We nw examine this me geneal questin. We begin with the stylized mdel f tw questins and then suggest a genealizatin t an abitay numbe f questins. 7

20 Utility Balance, Chice Balance, Cnfiguats, and M-Efficiency Stylized Analysis: Matching Questin Selectin t Manageial Fcus With metic questins (and pefeential independence) we can ask ne me independent questin, which we call a ange questin. We then cnside thee stylized pductdevelpment decisins. (Q4) p p = p p e ˆ ˆ Launch a single pduct The pduct-develpment manage has ne sht at the maket and has t chse the single best pduct t launch, pehaps because f shelf-space the limitatins. The manage is mst inteested in the difficult tadeffs, especially if the tw featues ae substitutes. The manage seeks pecisin n p p. Launch a pduct line (platfm decisin) The pduct-develpment manage is wking n a platfm fm which t launch multiple pducts, pehaps because f synegies in pductin. The manage is mst inteested in the ange f pducts t ffe (the beadth f the platfm). The manage seeks pecisin n p p. Launch a cnfiguat-based e-cmmece website The pduct-develpment manage plans t sell pducts with a web-based cnfiguat, pehaps because f flexible manufactuing capability. Hweve, thee is a cst t maintaining a bad inventy and thee is a cgnitive lad n the cnsume based n the cmplexity f the cnfiguat. The manage seeks pecisin n and n p. p F these stylized decisins we expect intuitively that we gain the mst pecisin by matching the questins t the manageial decisins. Table illustates this intuitin. F decisins n a single pduct launch, we gain the geatest pecisin by including tadeff questins (Q3); f decisins n pduct lines we gain the geatest pecisin by including ange questins (Q4), and f decisins n e-cmmece we gain the geatest pecisin by fcusing n featuespecific questins (Q and Q). These analytical esults, which match questin fmat t manageial need (f metic data), ae cmpaable in spiit t the empiical esults f Mashall and Badlw () wh examine cnguency between the fmats f the pfile and validatin data (metic vs. anking vs. chice) in hybid cnjint analysis. Mashall and Badlw s analyses suggest that pfile data equie less augmentatin with self-explicated data when they ae cnguent with the validatin data. Such questins ae feasible with a metic fmat. They wuld be eliminated as dminated altenatives in an dinal fmat such as stated-chice questins. 8

21 Utility Balance, Chice Balance, Cnfiguats, and M-Efficiency Table E Vaiances Due t Altenative Questin Selectin Tadeff Questins Range Questins Featue Fcus Single-pduct launch σ 3σ σ Pduct-line decisins 3σ σ σ e-cmmece website 3 σ 3 σ σ M-Efficiency The insights fm Table ae simple, yet pweful, and genealize eadily. In paticula, the stylized manageial decisins seek pecisin n Mp whee M is a linea tansfm f p. F example, if the manage is designing a cnfiguat and wants pecisin n the valuatin f featues elative t pice, then the manage wants t cncentate pecisin nt n the patwths pe se, but n specific cmbinatins f the patwths. We indicate this manageial fcus with a matix, M c, whee the last clumn cespnds t the pice patwth. This fcus matix pvides citeia, analgus t D-es and A-es, by which we can design questins. M-Efficiency f Metic Data the estimate f M c = We begin with metic data t illustate the citeia. F metic data, the standad e f Mp is pptinal t Σ = M ( X ' X ) M ' (Judge, et. al. 985, p. 57). By defining efficiency based n Σ M athe thanσ, we fcus pecisin n the patwths that matte mst t the manage s decisin. We call this fcus manageial efficiency (M-efficiency) and define M-es as fllws f suitably scaled X matices. M M D -e = q det( M ( X ' X ) M ' ) / n 9

22 Utility Balance, Chice Balance, Cnfiguats, and M-Efficiency M A -e = n M X X M tace q / ) ' ) ' ( ( The manageial fcus affects D-es and A-es diffeently. M D -e is defined by the deteminant f the mdified cvaiance matix and, hence, fcuses n the gemetic mean f the estimatin vaiances (eigenvalues). All else equal, the gemetic mean is minimized if the estimatin vaiances ae equal in magnitude. On the the hand, M A -e is defined by the tace f the cvaiance matix and, hence, fcuses n the aithmetic mean f the estimatin vaiances. Thus, if the manage wants diffeential fcus n sme cmbinatins f patwths and is willing t accept less pecisin n the cmbinatins, then M A -es might be a bette metic than M D - es. T illustate M-efficiency, we pvide an example in which we accept highe A- and D- es in de t educe M A - and M D -es. Cnside an thgnal aay, X. F this design, X X = I 6, whee I 6 is a 6x6 identify matix. The thgnal aay minimizes bth D-es and A-es. It des nt necessaily minimize eithe M D - M A -es as we illustate with the fivefeatue cnfiguat matix, M c. = c M = X

23 Utility Balance, Chice Balance, Cnfiguats, and M-Efficiency T incease M-efficiency, we petub X X M M clse t M cm c while maintaining full M M [ 6 c c ank. We chse XM such that X X = α βi ( β ) M M ] whee α and β ae scalas, β [,], and α is chsen t maintain cnstant questin magnitudes. 3 By cnstuctin, X M is n lnge an thgnal aay, hence bth A-es and D-es will incease. Hweve, M A - and M D -es decease because X M is me fcused n the manageial pblem. F example, with β = 4/5, M A -es decease by ve %; M D -es decease, but nt as damatically. Lage β s tun ut t be bette at educing M D -es, but less effective at educing M A -es. Deceases in M A -efficiency ae als pssible f squae M matices. Hweve, we cannt impve M D -es f squae M matices, because, f squae M, the questin design that minimizes D-e will als minimize M D -e. This is tue by the ppeties f the deteminant: det( Σ M ) = det(m(x X) - M ) = det(m)det((x X) - )det(m ) = det(m )det(m) det((x X) - ) = det(m M) det((x X) - ). As an illustatin, suppse the manage wants t fcus n cnfiguat pices (as in M c ), but als wants t fcus n the highest pice psted in the cnfiguat. T captue this fcus, we make M squae by adding anthe w t M c, [ -]. T decease M A - e we unbalance X t make pice me salient in the questin design. F example, a 4% unbalancing deceases M A -es by 5% with nly a slight incease in M D -es (<%). We chse these examples t be simple, yet illustative. We can ceate lage examples whee the deceases in M-es ae much lage. Ou petubatin algithms ae simple; me cmplex algithms might educe M-es futhe. M-efficiency appeas t be an inteesting and elevant citein n which t fcus futue algithmic develpment. M-Efficiency f Stated-Chice Questins The same aguments apply t stated-chice questins. The M-efficient citeia becme eithe the deteminant (M D -e) the tace (M A -e) f M(Z PZ) - M whee Z is a pbability-balanced X matix and P is a diagnal matix f chice pbabilities. See Aa and Hube (, p. 74) f ntatin. As in Sectin 4, thee will be a tadeff between chice balance and the levels f fcal featues. F M-es, that tadeff will be defined n cmbinatins f the featues athe than the featues themselves. 3 Recall that efficiency citeia tend t push featue levels t the extemes f the feasible space. Thus, t cmpae diffeent designs we must attempt t keep the magnitudes f the featues cnstant. In this example, we implement that cnstaint by keeping the tace f X M X M equal t the tace f X X. Ou gal is t pvide an example whee M- efficiency mattes. We can als ceate examples f the cnstaints the details wuld be diffeent, but the basic insights emain.

24 Utility Balance, Chice Balance, Cnfiguats, and M-Efficiency Summay f Sectins 5 and 6 Cnfiguat questins, based n analgies t e-cmmece websites, ae a elatively ecent fm f pefeence questins. Empiical data suggest that such questins ae supisingly gd at pedicting featue chice, in lage pat because they fcus pecisin n manageial issues that ae imptant f e-cmmece and mass custmizatin. This insight genealizes. Intuitively, pecisin (efficiency) is geatest when the questin fmat is matched t the manageial decisin. M A -es (and M D -es) pvide citeia by which match fmat t manageial elevance. These citeia can be used t mdify questin-design algithms. Thee emain the challenges f new algithmic develpment. Existing algithms have been ptimized and tested f D- and D p -efficiency. Hweve, Figue 4, Table, and the cncept f M-es have the ptential t efcus these algithms effectively. 7. Summay and Futue Reseach In the last few yeas geat stides have been made in develping algithms t select questins f bth metic paied-cmpaisn pefeence questins and f stated-chice questins. Algithms have been develped f a wide ange f estimatin methds including egessin, lgit, pbit, mixed lgit, and hieachical Bayes (bth metic and chice-based). In sme cases the algithms fcus n static designs and in the cases n adaptive designs. This existing eseach has led t pefeence questins that ae me efficient and me accuate as evidenced by bth simulatin and empiical expeiments. Summay In this pape, we use stylized mdels t gain insight n the ppeties f algithms and the citeia they ptimize. The mdels abstact fm the cmplexity f empiical estimatin t examine ppeties tanspaently. In key cases, the stylized mdels pvide insight n empiical facts. While sme f the fllwing insights fmalize that which was knwn fm pi algithmic eseach, many insights ae new and pvide a basis f futhe develpment: Metic utility balance leads t elative biases in patwth estimates. Metic utility balance is inefficient. Chice balance, while elated t utility balance, affects stated-chice questins diffeently than utility balance affects metic questins. When featues ae discete and the numbe f featues which can vay is fixed, chice balance is a cmpnent f ptimal efficiency.

25 Utility Balance, Chice Balance, Cnfiguats, and M-Efficiency Othewise, ptimal efficiency implies nn-ze chice balance. As espnse accuacy and/ the levels f nn-fcal featues incease, ptimal efficiency implies geate chice balance. Wam-up (cnfiguat) questins ae supisingly accuate f pedicting featue chice because they fcus pecisin n the featue-chice decisin. This cncept genealizes. Tadeff, ange, cnfiguat questins pvide geate efficiency when they ae matched t the apppiate manageial pblem. M-es pvide citeia by which bth metic and stated-chice questins can be fcused n the manageial decisins that will be made based n the data. Simple algithms (petubatin and/ unbalancing) can decease M-es effectively. The fmal mdels als pvide a they with which t undestand the accuacy f ecently ppsed plyhedal methds. Specifically, Metic plyhedal methds appea t avid the bias and inefficiency f utility balance by maintaining thgnality while fcusing questins whee patwths estimates ae less pecise. T date, stated-chice plyhedal methds have been applied f discete featues. Hence, thei use f appximate chice balance makes them clse t efficient. Hweve, stated-chice plyhedal methds f cntinuus featues might be impved with algithms that seek nn-ze chice balance. Futue Reseach This pape, with its fcus n stylized mdels, takes a diffeent tack than ecent algithmic papes. The tw tacks ae cmplementay. Ou stylized mdels can be extended t explain simulatin esults in Andews, Ainslie and Cuim (), empiical esults in Me (3), t exple the efficient use f self-explicated questins (Te Hfstede, im and Wedel ). On the the hand, insights n utility balance, chice balance, and M-efficiency can be used t impve algithms f bth aggegate custmizatin and plyhedal methds. Finally, inteest in cnfiguat-like questins is gwing as me maketing eseach mves t the web and as e-cmmece manages demand maketing eseach that matches the decisins they make daily. We have nly begun t undestand the natue f these questins. 3

26 Utility Balance, Chice Balance, Cnfiguats, and M-Efficiency, Refeences Refeences Andews, Rick L., Andew Ainslie, and Iman S. Cuim (), An Empiical Cmpaisn f Lgit Chice Mdels with Discete vesus Cntinuus Repesentatins f Hetegeneity, Junal f Maketing Reseach, 39, (Nvembe), Aa, Neeaj and Jel Hube (), Impving Paamete Estimates and Mdel Pedictin by Aggegate Custmizatin in Chice Expeiments, Junal f Cnsume Reseach, 8, (Septembe), Dahan, Ely and Jhn R. Hause (), "The Vitual Custme," Junal f Pduct Innvatin Management, 9, 5, (Septembe), Delquie, Philippe (3), Optimal Cnflict in Pefeence Assessment, Management Science, 49,, (Januay), -5. Evgeniu, Theds, Cnstantins Bussis, and Gigs Zachaia (3), Genealized Rbust Cnjint Estimatin, Wking Pape, (Fntainebleau, Fance: INSEAD), May. Eld, Tey, Jdan Luviee, and ishnakuma S. Davey (99), An Empiical Cmpaisn f Ratings-Based and Chice-based Cnjint Mdels, Junal f Maketing Reseach 9, 3, (August), Geen, Paul E, Abba iege, and Manj. Agawal (99), Adaptive Cnjint Analysis: Sme Caveats and Suggestins, Junal f Maketing Reseach, pp. 5-.,, and Yam Wind (), Thity Yeas f Cnjint Analysis: Reflectins and Pspects, Intefaces, 3, 3, Pat, (May-June), S56-S73. Geene, William H. (993), Ecnmetic Analysis, E, (Englewd Cliffs, NJ: Pentice-Hall, Inc.) Haaije, Rinus, Wagne amakua, and Michel Wedel (), Respnse Latencies in the Analysis f Cnjint Chice Expeiments, Junal f Maketing Reseach 37, (August), Hause, Jhn R. and Steven M. Shugan (98), Intensity Measues f Cnsume Pefeence, Opeatins Reseach, 8,, (Mach-Apil), Hube, Jel and David Hansen (986), Testing the Impact f Dimensinal Cmplexity and Affective Diffeences f Paied Cncepts in Adaptive Cnjint Analysis, Advances in Cnsume Reseach, 4, Melanie Wallendf and Paul Andesn, eds., Pv, UT: Assciatins f Cnsume Reseach, and laus Zweina (996), The Imptance f Utility Balance in Efficient Chice Designs, Junal f Maketing Reseach, 33, (August), R

OBJECTIVE To investigate the parallel connection of R, L, and C. 1 Electricity & Electronics Constructor EEC470

OBJECTIVE To investigate the parallel connection of R, L, and C. 1 Electricity & Electronics Constructor EEC470 Assignment 7 Paallel Resnance OBJECTIVE T investigate the paallel cnnectin f R,, and C. EQUIPMENT REQUIRED Qty Appaatus 1 Electicity & Electnics Cnstuct EEC470 1 Basic Electicity and Electnics Kit EEC471-1

More information

ME 3600 Control Systems Frequency Domain Analysis

ME 3600 Control Systems Frequency Domain Analysis ME 3600 Cntl Systems Fequency Dmain Analysis The fequency espnse f a system is defined as the steady-state espnse f the system t a sinusidal (hamnic) input. F linea systems, the esulting utput is itself

More information

The Gradient and Applications This unit is based on Sections 9.5 and 9.6, Chapter 9. All assigned readings and exercises are from the textbook

The Gradient and Applications This unit is based on Sections 9.5 and 9.6, Chapter 9. All assigned readings and exercises are from the textbook The Gadient and Applicatins This unit is based n Sectins 9.5 and 9.6 Chapte 9. All assigned eadings and eecises ae fm the tetbk Objectives: Make cetain that u can define and use in cntet the tems cncepts

More information

CHAPTER 24 GAUSS LAW

CHAPTER 24 GAUSS LAW CHAPTR 4 GAUSS LAW LCTRIC FLUX lectic flux is a measue f the numbe f electic filed lines penetating sme suface in a diectin pependicula t that suface. Φ = A = A csθ with θ is the angle between the and

More information

MEM202 Engineering Mechanics Statics Course Web site:

MEM202 Engineering Mechanics Statics Course Web site: 0 Engineeing Mechanics - Statics 0 Engineeing Mechanics Statics Cuse Web site: www.pages.dexel.edu/~cac54 COUSE DESCIPTION This cuse cves intemediate static mechanics, an extensin f the fundamental cncepts

More information

Journal of Theoretics

Journal of Theoretics Junal f Theetics Junal Hme Page The Classical Pblem f a Bdy Falling in a Tube Thugh the Cente f the Eath in the Dynamic They f Gavity Iannis Iaklis Haanas Yk Univesity Depatment f Physics and Astnmy A

More information

Analytical Solution to Diffusion-Advection Equation in Spherical Coordinate Based on the Fundamental Bloch NMR Flow Equations

Analytical Solution to Diffusion-Advection Equation in Spherical Coordinate Based on the Fundamental Bloch NMR Flow Equations Intenatinal Junal f heetical and athematical Phsics 5, 5(5: 4-44 OI:.593/j.ijtmp.555.7 Analtical Slutin t iffusin-advectin Equatin in Spheical Cdinate Based n the Fundamental Blch N Flw Equatins anladi

More information

Distance oracles in edge-labeled graphs

Distance oracles in edge-labeled graphs Distance acles in edge-labeled gaphs Fancesc Bnchi Aistides Ginis Fancesc Gull Antti Ukknen Yah Labs Bacelna, Spain {bnchi,gull}@yah-inc.cm Helsinki Institute f Infmatin Technlgy HIIT Aalt Univesity {aistides.ginis,antti.ukknen}@aalt.fi

More information

5.1 Moment of a Force Scalar Formation

5.1 Moment of a Force Scalar Formation Outline ment f a Cuple Equivalent System Resultants f a Fce and Cuple System ment f a fce abut a pint axis a measue f the tendency f the fce t cause a bdy t tate abut the pint axis Case 1 Cnside hizntal

More information

Introduction. Electrostatics

Introduction. Electrostatics UNIVESITY OF TECHNOLOGY, SYDNEY FACULTY OF ENGINEEING 4853 Electmechanical Systems Electstatics Tpics t cve:. Culmb's Law 5. Mateial Ppeties. Electic Field Stength 6. Gauss' Theem 3. Electic Ptential 7.

More information

Optimal Design of Transonic Fan Blade Leading Edge Shape Using CFD and Simultaneous Perturbation Stochastic Approximation Method

Optimal Design of Transonic Fan Blade Leading Edge Shape Using CFD and Simultaneous Perturbation Stochastic Approximation Method 1 Optimal Design f Tansnic Fan Blade Leading Edge Shape Using CFD and Simultaneus Petubatin Stchastic Appximatin Methd X.Q.Xing and M.Damdaan Abstact Simultaneus Petubatin Stchastic Appximatin methd has

More information

Phys 332 Electricity & Magnetism Day 3. Note: I should have recommended reading section 1.5 (delta function) as well. rˆ rˆ

Phys 332 Electricity & Magnetism Day 3. Note: I should have recommended reading section 1.5 (delta function) as well. rˆ rˆ Phs 33 lecticit & Magnetism Da 3 Mn. 9/9 Wed. 9/ Thus 9/ Fi. 9/3 (C.-.5,.8). &.5;..-.. Gauss & Div, T Numeical Quadatue (C.-.5,.8)..3 Using Gauss (C.-.5,.8)..3-.. Using Gauss HW quipment Bing in ppt s

More information

EM Boundary Value Problems

EM Boundary Value Problems EM Bounday Value Poblems 10/ 9 11/ By Ilekta chistidi & Lee, Seung-Hyun A. Geneal Desciption : Maxwell Equations & Loentz Foce We want to find the equations of motion of chaged paticles. The way to do

More information

Subjects discussed: Aircraft Engine Noise : Principles; Regulations

Subjects discussed: Aircraft Engine Noise : Principles; Regulations 16.50 Lectue 36 Subjects discussed: Aicaft Engine Nise : Pinciples; Regulatins Nise geneatin in the neighbhds f busy aipts has been a seius pblem since the advent f the jet-pweed tanspt, in the late 1950's.

More information

INVERSE QUANTUM STATES OF HYDROGEN

INVERSE QUANTUM STATES OF HYDROGEN INVERSE QUANTUM STATES OF HYDROGEN Rnald C. Bugin Edgecmbe Cmmunity Cllege Rcky Munt, Nth Calina 780 bugin@edgecmbe.edu ABSTRACT The pssible existence f factinal quantum states in the hydgen atm has been

More information

Consider the simple circuit of Figure 1 in which a load impedance of r is connected to a voltage source. The no load voltage of r

Consider the simple circuit of Figure 1 in which a load impedance of r is connected to a voltage source. The no load voltage of r 1 Intductin t Pe Unit Calculatins Cnside the simple cicuit f Figue 1 in which a lad impedance f L 60 + j70 Ω 9. 49 Ω is cnnected t a vltage suce. The n lad vltage f the suce is E 1000 0. The intenal esistance

More information

School of Chemical & Biological Engineering, Konkuk University

School of Chemical & Biological Engineering, Konkuk University Schl f Cheical & Bilgical Engineeing, Knkuk Univesity Lectue 7 Ch. 2 The Fist Law Thecheisty Pf. Y-Sep Min Physical Cheisty I, Sping 2008 Ch. 2-2 The study f the enegy tansfeed as heat duing the cuse f

More information

Outline. Steady Heat Transfer with Conduction and Convection. Review Steady, 1-D, Review Heat Generation. Review Heat Generation II

Outline. Steady Heat Transfer with Conduction and Convection. Review Steady, 1-D, Review Heat Generation. Review Heat Generation II Steady Heat ansfe ebuay, 7 Steady Heat ansfe wit Cnductin and Cnvectin ay Caett Mecanical Engineeing 375 Heat ansfe ebuay, 7 Outline eview last lectue Equivalent cicuit analyses eview basic cncept pplicatin

More information

Microelectronics Circuit Analysis and Design. ac Equivalent Circuit for Common Emitter. Common Emitter with Time-Varying Input

Microelectronics Circuit Analysis and Design. ac Equivalent Circuit for Common Emitter. Common Emitter with Time-Varying Input Micelectnics Cicuit Analysis and Design Dnald A. Neamen Chapte 6 Basic BJT Amplifies In this chapte, we will: Undestand the pinciple f a linea amplifie. Discuss and cmpae the thee basic tansist amplifie

More information

Empirical Prediction of Fitting Densities in Industrial Workrooms for Ray Tracing. 1 Introduction. 2 Ray Tracing using DRAYCUB

Empirical Prediction of Fitting Densities in Industrial Workrooms for Ray Tracing. 1 Introduction. 2 Ray Tracing using DRAYCUB Empiical Pediction of Fitting Densities in Industial Wokooms fo Ray Tacing Katina Scheebnyj, Muay Hodgson Univesity of Bitish Columbia, SOEH-MECH, Acoustics and Noise Reseach Goup, 226 East Mall, Vancouve,

More information

Research Design - - Topic 17 Multiple Regression & Multiple Correlation: Two Predictors 2009 R.C. Gardner, Ph.D.

Research Design - - Topic 17 Multiple Regression & Multiple Correlation: Two Predictors 2009 R.C. Gardner, Ph.D. Reseach Design - - Topic 7 Multiple Regession & Multiple Coelation: Two Pedictos 009 R.C. Gadne, Ph.D. Geneal Rationale and Basic Aithmetic fo two pedictos Patial and semipatial coelation Regession coefficients

More information

Steady State Analysis of Squirrel-Cage Induction Machine with Skin-Effect

Steady State Analysis of Squirrel-Cage Induction Machine with Skin-Effect Steady State Analysis f Squiel-Cage Inductin Machine with Skin-Effect D.-Ing. O. I. Ok Depatment f Electical Engineeing Univesity f Nigeia, Nsukka Enugu State, Nigeia. Email: gnnayak@htmail.cm ABSTACT

More information

Bootstrap Method > # Purpose: understand how bootstrap method works > obs=c(11.96, 5.03, 67.40, 16.07, 31.50, 7.73, 11.10, 22.38) > n=length(obs) >

Bootstrap Method > # Purpose: understand how bootstrap method works > obs=c(11.96, 5.03, 67.40, 16.07, 31.50, 7.73, 11.10, 22.38) > n=length(obs) > Btstrap Methd > # Purpse: understand hw btstrap methd wrks > bs=c(11.96, 5.03, 67.40, 16.07, 31.50, 7.73, 11.10, 22.38) > n=length(bs) > mean(bs) [1] 21.64625 > # estimate f lambda > lambda = 1/mean(bs);

More information

MEASURING CHINESE RISK AVERSION

MEASURING CHINESE RISK AVERSION MEASURING CHINESE RISK AVERSION --Based on Insuance Data Li Diao (Cental Univesity of Finance and Economics) Hua Chen (Cental Univesity of Finance and Economics) Jingzhen Liu (Cental Univesity of Finance

More information

6 PROBABILITY GENERATING FUNCTIONS

6 PROBABILITY GENERATING FUNCTIONS 6 PROBABILITY GENERATING FUNCTIONS Cetain deivations pesented in this couse have been somewhat heavy on algeba. Fo example, detemining the expectation of the Binomial distibution (page 5.1 tuned out to

More information

On the Micropolar Fluid Flow through Porous Media

On the Micropolar Fluid Flow through Porous Media Pceedings f the th WEA Int. Cnf. n MATHEMATICAL METHOD, COMPUTATIONAL TECHNIQUE AND INTELLIGENT YTEM On the Micpla Fluid Flw thugh Pus Media M.T. KAMEL 3, D. ROACH, M.H. HAMDAN,3 Depatment f Mathematical

More information

Determining solar characteristics using planetary data

Determining solar characteristics using planetary data Detemining sola chaacteistics using planetay data Intoduction The Sun is a G-type main sequence sta at the cente of the Sola System aound which the planets, including ou Eath, obit. In this investigation

More information

Chapter 4 Motion in Two and Three Dimensions

Chapter 4 Motion in Two and Three Dimensions Chapte 4 Mtin in Tw and Thee Dimensins In this chapte we will cntinue t stud the mtin f bjects withut the estictin we put in chapte t me aln a staiht line. Instead we will cnside mtin in a plane (tw dimensinal

More information

R/2 L/2 i + di ---> + + v G C. v + dv - - <--- i. R, L, G, C per unit length

R/2 L/2 i + di ---> + + v G C. v + dv - - <--- i. R, L, G, C per unit length _03_EE394J_2_Sping2_Tansmissin_Linesdc Tansmissin Lines Inductance and capacitance calculatins f tansmissin lines GMR GM L and C matices effect f gund cnductivity Undegund cables Equivalent Cicuit f Tansmissin

More information

Do Managers Do Good With Other People s Money? Online Appendix

Do Managers Do Good With Other People s Money? Online Appendix Do Manages Do Good With Othe People s Money? Online Appendix Ing-Haw Cheng Haison Hong Kelly Shue Abstact This is the Online Appendix fo Cheng, Hong and Shue 2013) containing details of the model. Datmouth

More information

Basic Bridge Circuits

Basic Bridge Circuits AN7 Datafoth Copoation Page of 6 DID YOU KNOW? Samuel Hunte Chistie (784-865) was bon in London the son of James Chistie, who founded Chistie's Fine At Auctionees. Samuel studied mathematics at Tinity

More information

B. Spherical Wave Propagation

B. Spherical Wave Propagation 11/8/007 Spheical Wave Popagation notes 1/1 B. Spheical Wave Popagation Evey antenna launches a spheical wave, thus its powe density educes as a function of 1, whee is the distance fom the antenna. We

More information

Department of Economics, University of California, Davis Ecn 200C Micro Theory Professor Giacomo Bonanno. Insurance Markets

Department of Economics, University of California, Davis Ecn 200C Micro Theory Professor Giacomo Bonanno. Insurance Markets Department f Ecnmics, University f alifrnia, Davis Ecn 200 Micr Thery Prfessr Giacm Bnann Insurance Markets nsider an individual wh has an initial wealth f. ith sme prbability p he faces a lss f x (0

More information

n 1 Cov(X,Y)= ( X i- X )( Y i-y ). N-1 i=1 * If variable X and variable Y tend to increase together, then c(x,y) > 0

n 1 Cov(X,Y)= ( X i- X )( Y i-y ). N-1 i=1 * If variable X and variable Y tend to increase together, then c(x,y) > 0 Covaiance and Peason Coelation Vatanian, SW 540 Both covaiance and coelation indicate the elationship between two (o moe) vaiables. Neithe the covaiance o coelation give the slope between the X and Y vaiable,

More information

Electromagnetic Waves

Electromagnetic Waves Chapte 3 lectmagnetic Waves 3.1 Maxwell s quatins and ectmagnetic Waves A. Gauss s Law: # clsed suface aea " da Q enc lectic fields may be geneated by electic chages. lectic field lines stat at psitive

More information

MATH 415, WEEK 3: Parameter-Dependence and Bifurcations

MATH 415, WEEK 3: Parameter-Dependence and Bifurcations MATH 415, WEEK 3: Paamete-Dependence and Bifucations 1 A Note on Paamete Dependence We should pause to make a bief note about the ole played in the study of dynamical systems by the system s paametes.

More information

Psychometric Methods: Theory into Practice Larry R. Price

Psychometric Methods: Theory into Practice Larry R. Price ERRATA Psychometic Methods: Theoy into Pactice Lay R. Pice Eos wee made in Equations 3.5a and 3.5b, Figue 3., equations and text on pages 76 80, and Table 9.1. Vesions of the elevant pages that include

More information

The piezoelectric quartz crystal resonator is an essential

The piezoelectric quartz crystal resonator is an essential 738 IEEE Tansactins n Ultasnics, Feelectics, and Fequency Cntl, vl. 59, n. 4, Apil 0 Vltage-Cntlled Duble-Resnance Quatz Oscillat Using Vaiable-Capacitance Dide Ruzaini Izyan Binti Ruslan, Tmi Sath, and

More information

PERFORMANCE OPTIMIZATION OF LINEAR INDUCTION MOTOR BY EDDY CURRENT AND FLUX DENSITY DISTRIBUTION ANALYSIS

PERFORMANCE OPTIMIZATION OF LINEAR INDUCTION MOTOR BY EDDY CURRENT AND FLUX DENSITY DISTRIBUTION ANALYSIS unal f Engineeing Science and Technlg Vl. 6, N. 6 (011) 769-776 Schl f Engineeing, Tal s Univesit PERFORMANCE OPTIMIZATION OF LINEAR INDUCTION MOTOR BY EDDY CURRENT AND FLUX DENSITY DISTRIBUTION ANALYSIS

More information

A Crash Course in (2 2) Matrices

A Crash Course in (2 2) Matrices A Cash Couse in ( ) Matices Seveal weeks woth of matix algeba in an hou (Relax, we will only stuy the simplest case, that of matices) Review topics: What is a matix (pl matices)? A matix is a ectangula

More information

HOW TO TEACH THE FUNDAMENTALS OF INFORMATION SCIENCE, CODING, DECODING AND NUMBER SYSTEMS?

HOW TO TEACH THE FUNDAMENTALS OF INFORMATION SCIENCE, CODING, DECODING AND NUMBER SYSTEMS? 6th INTERNATIONAL MULTIDISCIPLINARY CONFERENCE HOW TO TEACH THE FUNDAMENTALS OF INFORMATION SCIENCE, CODING, DECODING AND NUMBER SYSTEMS? Cecília Sitkuné Göömbei College of Nyíegyháza Hungay Abstact: The

More information

Boundary Layers and Singular Perturbation Lectures 16 and 17 Boundary Layers and Singular Perturbation. x% 0 Ω0æ + Kx% 1 Ω0æ + ` : 0. (9.

Boundary Layers and Singular Perturbation Lectures 16 and 17 Boundary Layers and Singular Perturbation. x% 0 Ω0æ + Kx% 1 Ω0æ + ` : 0. (9. Lectues 16 and 17 Bounday Layes and Singula Petubation A Regula Petubation In some physical poblems, the solution is dependent on a paamete K. When the paamete K is vey small, it is natual to expect that

More information

ASTR415: Problem Set #6

ASTR415: Problem Set #6 ASTR45: Poblem Set #6 Cuan D. Muhlbege Univesity of Mayland (Dated: May 7, 27) Using existing implementations of the leapfog and Runge-Kutta methods fo solving coupled odinay diffeential equations, seveal

More information

CRESTMORE SKY BLUE MARBLE, ITS LINEAR THERMAL EXPANSION AND COLOR

CRESTMORE SKY BLUE MARBLE, ITS LINEAR THERMAL EXPANSION AND COLOR CRESTMORE SKY BLE MARBLE, ITS LINEAR THERMAL EXPANSION AND COLOR Jsnen L. RsBxnlz AND Duny T. Sun, Rensselae Pl^'ttechnic Institute. Tt. Nezp Yk. Aslc Themal studies wee made f the Sky Blue mable fm Cestme,

More information

Chapter 2: Estuarine Salinity Structure and Circulation

Chapter 2: Estuarine Salinity Structure and Circulation Chapte : Estuaine Salinity Stuctue and Ciculatin W.R. Geye, Wds Hle Oceangaphic Institutin.. The Hizntal Salinity Gadient Estuaies shw a geat divesity f size, shape, depth, and fcing chaacteistics, but

More information

Solution to Problem First, the firm minimizes the cost of the inputs: min wl + rk + sf

Solution to Problem First, the firm minimizes the cost of the inputs: min wl + rk + sf Econ 0A Poblem Set 4 Solutions ue in class on Tu 4 Novembe. No late Poblem Sets accepted, so! This Poblem set tests the knoledge that ou accumulated mainl in lectues 5 to 9. Some of the mateial ill onl

More information

Introduction to Nuclear Forces

Introduction to Nuclear Forces Intoduction to Nuclea Foces One of the main poblems of nuclea physics is to find out the natue of nuclea foces. Nuclea foces diffe fom all othe known types of foces. They cannot be of electical oigin since

More information

Safety variations in steel designed using Eurocode 3

Safety variations in steel designed using Eurocode 3 JCSS Wokshop on eliability Based Code Calibation Safety vaiations in steel designed using Euocode 3 Mike Byfield Canfield Univesity Swindon, SN6 8LA, UK David Nethecot Impeial College London SW7 2BU, UK

More information

Section 8.2 Polar Coordinates

Section 8.2 Polar Coordinates Section 8. Pola Coodinates 467 Section 8. Pola Coodinates The coodinate system we ae most familia with is called the Catesian coodinate system, a ectangula plane divided into fou quadants by the hoizontal

More information

2. Electrostatics. Dr. Rakhesh Singh Kshetrimayum 8/11/ Electromagnetic Field Theory by R. S. Kshetrimayum

2. Electrostatics. Dr. Rakhesh Singh Kshetrimayum 8/11/ Electromagnetic Field Theory by R. S. Kshetrimayum 2. Electostatics D. Rakhesh Singh Kshetimayum 1 2.1 Intoduction In this chapte, we will study how to find the electostatic fields fo vaious cases? fo symmetic known chage distibution fo un-symmetic known

More information

An Inventory Model for Two Warehouses with Constant Deterioration and Quadratic Demand Rate under Inflation and Permissible Delay in Payments

An Inventory Model for Two Warehouses with Constant Deterioration and Quadratic Demand Rate under Inflation and Permissible Delay in Payments Ameican Jounal of Engineeing Reseach (AJER) 16 Ameican Jounal of Engineeing Reseach (AJER) e-issn: 3-847 p-issn : 3-936 Volume-5, Issue-6, pp-6-73 www.aje.og Reseach Pape Open Access An Inventoy Model

More information

COMPUTATIONS OF ELECTROMAGNETIC FIELDS RADIATED FROM COMPLEX LIGHTNING CHANNELS

COMPUTATIONS OF ELECTROMAGNETIC FIELDS RADIATED FROM COMPLEX LIGHTNING CHANNELS Pogess In Electomagnetics Reseach, PIER 73, 93 105, 2007 COMPUTATIONS OF ELECTROMAGNETIC FIELDS RADIATED FROM COMPLEX LIGHTNING CHANNELS T.-X. Song, Y.-H. Liu, and J.-M. Xiong School of Mechanical Engineeing

More information

7.2. Coulomb s Law. The Electric Force

7.2. Coulomb s Law. The Electric Force Coulomb s aw Recall that chaged objects attact some objects and epel othes at a distance, without making any contact with those objects Electic foce,, o the foce acting between two chaged objects, is somewhat

More information

Motion in One Dimension

Motion in One Dimension Motion in One Dimension Intoduction: In this lab, you will investigate the motion of a olling cat as it tavels in a staight line. Although this setup may seem ovesimplified, you will soon see that a detailed

More information

equilibrium in the money market

equilibrium in the money market Sahoko KAJI --- Open Economy Macoeconomics ectue Notes I I A Review of Closed Economy Macoeconomics We begin by eviewing some of the basics of closed economy macoeconomics that ae indispensable in undestanding

More information

Current Balance Warm Up

Current Balance Warm Up PHYSICS EXPERIMENTS 133 Cuent Balance-1 Cuent Balance Wam Up 1. Foce between cuent-caying wies Wie 1 has a length L (whee L is "long") and caies a cuent I 0. What is the magnitude of the magnetic field

More information

Lyapunov Stability Stability of Equilibrium Points

Lyapunov Stability Stability of Equilibrium Points Lyapunv Stability Stability f Equilibrium Pints 1. Stability f Equilibrium Pints - Definitins In this sectin we cnsider n-th rder nnlinear time varying cntinuus time (C) systems f the frm x = f ( t, x),

More information

On the Quasi-inverse of a Non-square Matrix: An Infinite Solution

On the Quasi-inverse of a Non-square Matrix: An Infinite Solution Applied Mathematical Sciences, Vol 11, 2017, no 27, 1337-1351 HIKARI Ltd, wwwm-hikaicom https://doiog/1012988/ams20177273 On the Quasi-invese of a Non-squae Matix: An Infinite Solution Ruben D Codeo J

More information

ON THE K-FACTOR DISTRIBUTION AND DIFFRACTION FADING FOR SOUTHERN AFRICA

ON THE K-FACTOR DISTRIBUTION AND DIFFRACTION FADING FOR SOUTHERN AFRICA 7 SOUTH AFRICAN INSTITUTE OF ELECTRICAL ENGINEERS Vl.97() June 006 ON THE K-FACTOR DISTRIBUTION AND DIFFRACTION AND DIFFRACTION FADING FOR SOUTHERN AFRICA T.J. Afull and P.K. Odedina Schl f Electical,

More information

Ch. 3: Inverse Kinematics Ch. 4: Velocity Kinematics. The Interventional Centre

Ch. 3: Inverse Kinematics Ch. 4: Velocity Kinematics. The Interventional Centre Ch. : Invee Kinemati Ch. : Velity Kinemati The Inteventinal Cente eap: kinemati eupling Apppiate f ytem that have an am a wit Suh that the wit jint ae ae aligne at a pint F uh ytem, we an plit the invee

More information

Lead field theory and the spatial sensitivity of scalp EEG Thomas Ferree and Matthew Clay July 12, 2000

Lead field theory and the spatial sensitivity of scalp EEG Thomas Ferree and Matthew Clay July 12, 2000 Lead field theoy and the spatial sensitivity of scalp EEG Thomas Feee and Matthew Clay July 12, 2000 Intoduction Neuonal population activity in the human cotex geneates electic fields which ae measuable

More information

APPLICATION OF MAC IN THE FREQUENCY DOMAIN

APPLICATION OF MAC IN THE FREQUENCY DOMAIN PPLICION OF MC IN HE FREQUENCY DOMIN D. Fotsch and D. J. Ewins Dynamics Section, Mechanical Engineeing Depatment Impeial College of Science, echnology and Medicine London SW7 2B, United Kingdom BSRC he

More information

When two numbers are written as the product of their prime factors, they are in factored form.

When two numbers are written as the product of their prime factors, they are in factored form. 10 1 Study Guide Pages 420 425 Factos Because 3 4 12, we say that 3 and 4 ae factos of 12. In othe wods, factos ae the numbes you multiply to get a poduct. Since 2 6 12, 2 and 6 ae also factos of 12. The

More information

Lecture 4. Electric Potential

Lecture 4. Electric Potential Lectue 4 Electic Ptentil In this lectue yu will len: Electic Scl Ptentil Lplce s n Pissn s Eutin Ptentil f Sme Simple Chge Distibutins ECE 0 Fll 006 Fhn Rn Cnell Univesity Cnsevtive Ittinl Fiels Ittinl

More information

CENTER FOR MULTIMODAL SOLUTIONS FOR CONGESTION MITIGATION (CMS)

CENTER FOR MULTIMODAL SOLUTIONS FOR CONGESTION MITIGATION (CMS) Final Repot to the CENTER FOR MULTIMODAL SOLUTIONS FOR CONGESTION MITIGATION (CMS) CMS Poect Numbe: _8-4_ Title: Chaacteizing the Tadeoffs and Costs Associated with Tanspotation Congestion in Supply Chains

More information

Likelihood vs. Information in Aligning Biopolymer Sequences. UCSD Technical Report CS Timothy L. Bailey

Likelihood vs. Information in Aligning Biopolymer Sequences. UCSD Technical Report CS Timothy L. Bailey Likelihood vs. Infomation in Aligning Biopolyme Sequences UCSD Technical Repot CS93-318 Timothy L. Bailey Depatment of Compute Science and Engineeing Univesity of Califonia, San Diego 1 Febuay, 1993 ABSTRACT:

More information

6 Matrix Concentration Bounds

6 Matrix Concentration Bounds 6 Matix Concentation Bounds Concentation bounds ae inequalities that bound pobabilities of deviations by a andom vaiable fom some value, often its mean. Infomally, they show the pobability that a andom

More information

Right-handed screw dislocation in an isotropic solid

Right-handed screw dislocation in an isotropic solid Dislocation Mechanics Elastic Popeties of Isolated Dislocations Ou study of dislocations to this point has focused on thei geomety and thei ole in accommodating plastic defomation though thei motion. We

More information

Many Electron Atoms. Electrons can be put into approximate orbitals and the properties of the many electron systems can be catalogued

Many Electron Atoms. Electrons can be put into approximate orbitals and the properties of the many electron systems can be catalogued Many Electon Atoms The many body poblem cannot be solved analytically. We content ouselves with developing appoximate methods that can yield quite accuate esults (but usually equie a compute). The electons

More information

Lead/Lag Compensator Frequency Domain Properties and Design Methods

Lead/Lag Compensator Frequency Domain Properties and Design Methods Lectures 6 and 7 Lead/Lag Cmpensatr Frequency Dmain Prperties and Design Methds Definitin Cnsider the cmpensatr (ie cntrller Fr, it is called a lag cmpensatr s K Fr s, it is called a lead cmpensatr Ntatin

More information

In statistical computations it is desirable to have a simplified system of notation to avoid complicated formulas describing mathematical operations.

In statistical computations it is desirable to have a simplified system of notation to avoid complicated formulas describing mathematical operations. Chapte 1 STATISTICAL NOTATION AND ORGANIZATION 11 Summation Notation fo a One-Way Classification In statistical computations it is desiable to have a simplified system of notation to avoid complicated

More information

A Few Basic Facts About Isothermal Mass Transfer in a Binary Mixture

A Few Basic Facts About Isothermal Mass Transfer in a Binary Mixture Few asic Facts but Isthermal Mass Transfer in a inary Miture David Keffer Department f Chemical Engineering University f Tennessee first begun: pril 22, 2004 last updated: January 13, 2006 dkeffer@utk.edu

More information

A Comparison and Contrast of Some Methods for Sample Quartiles

A Comparison and Contrast of Some Methods for Sample Quartiles A Compaison and Contast of Some Methods fo Sample Quatiles Anwa H. Joade and aja M. Latif King Fahd Univesity of Petoleum & Mineals ABSTACT A emainde epesentation of the sample size n = 4m ( =, 1, 2, 3)

More information

February 28, 2013 COMMENTS ON DIFFUSION, DIFFUSIVITY AND DERIVATION OF HYPERBOLIC EQUATIONS DESCRIBING THE DIFFUSION PHENOMENA

February 28, 2013 COMMENTS ON DIFFUSION, DIFFUSIVITY AND DERIVATION OF HYPERBOLIC EQUATIONS DESCRIBING THE DIFFUSION PHENOMENA February 28, 2013 COMMENTS ON DIFFUSION, DIFFUSIVITY AND DERIVATION OF HYPERBOLIC EQUATIONS DESCRIBING THE DIFFUSION PHENOMENA Mental Experiment regarding 1D randm walk Cnsider a cntainer f gas in thermal

More information

30 The Electric Field Due to a Continuous Distribution of Charge on a Line

30 The Electric Field Due to a Continuous Distribution of Charge on a Line hapte 0 The Electic Field Due to a ontinuous Distibution of hage on a Line 0 The Electic Field Due to a ontinuous Distibution of hage on a Line Evey integal ust include a diffeential (such as d, dt, dq,

More information

A) 100 K B) 150 K C) 200 K D) 250 K E) 350 K

A) 100 K B) 150 K C) 200 K D) 250 K E) 350 K Phys10 Secnd Maj-09 Ze Vesin Cdinat: k Wednesday, May 05, 010 Page: 1 Q1. A ht bject and a cld bject ae placed in themal cntact and the cmbinatin is islated. They tansfe enegy until they each a final equilibium

More information

THE JEU DE TAQUIN ON THE SHIFTED RIM HOOK TABLEAUX. Jaejin Lee

THE JEU DE TAQUIN ON THE SHIFTED RIM HOOK TABLEAUX. Jaejin Lee Koean J. Math. 23 (2015), No. 3, pp. 427 438 http://dx.doi.og/10.11568/kjm.2015.23.3.427 THE JEU DE TAQUIN ON THE SHIFTED RIM HOOK TABLEAUX Jaejin Lee Abstact. The Schensted algoithm fist descibed by Robinson

More information

Lab #4: Newton s Second Law

Lab #4: Newton s Second Law Lab #4: Newton s Second Law Si Isaac Newton Reading Assignment: bon: Januay 4, 1643 Chapte 5 died: Mach 31, 1727 Chapte 9, Section 9-7 Intoduction: Potait of Isaac Newton by Si Godfey Knelle http://www.newton.cam.ac.uk/at/potait.html

More information

3.6 Applied Optimization

3.6 Applied Optimization .6 Applied Optimization Section.6 Notes Page In this section we will be looking at wod poblems whee it asks us to maimize o minimize something. Fo all the poblems in this section you will be taking the

More information

Newton s Laws, Kepler s Laws, and Planetary Orbits

Newton s Laws, Kepler s Laws, and Planetary Orbits Newton s Laws, Keple s Laws, and Planetay Obits PROBLEM SET 4 DUE TUESDAY AT START OF LECTURE 28 Septembe 2017 ASTRONOMY 111 FALL 2017 1 Newton s & Keple s laws and planetay obits Unifom cicula motion

More information

Light Time Delay and Apparent Position

Light Time Delay and Apparent Position Light Time Delay and ppaent Position nalytical Gaphics, Inc. www.agi.com info@agi.com 610.981.8000 800.220.4785 Contents Intoduction... 3 Computing Light Time Delay... 3 Tansmission fom to... 4 Reception

More information

This is a very simple sampling mode, and this article propose an algorithm about how to recover x from y in this condition.

This is a very simple sampling mode, and this article propose an algorithm about how to recover x from y in this condition. 3d Intenational Confeence on Multimedia echnology(icm 03) A Simple Compessive Sampling Mode and the Recovey of Natue Images Based on Pixel Value Substitution Wenping Shao, Lin Ni Abstact: Compessive Sampling

More information

3.2 Centripetal Acceleration

3.2 Centripetal Acceleration unifom cicula motion the motion of an object with onstant speed along a cicula path of constant adius 3.2 Centipetal Acceleation The hamme thow is a tack-and-field event in which an athlete thows a hamme

More information

Magnetic Field. Conference 6. Physics 102 General Physics II

Magnetic Field. Conference 6. Physics 102 General Physics II Physics 102 Confeence 6 Magnetic Field Confeence 6 Physics 102 Geneal Physics II Monday, Mach 3d, 2014 6.1 Quiz Poblem 6.1 Think about the magnetic field associated with an infinite, cuent caying wie.

More information

MASSACHUSETTS INSTITUTE OF TECHNOLOGY Physics Department Physics 8.07: Electromagnetism II September 15, 2012 Prof. Alan Guth PROBLEM SET 2

MASSACHUSETTS INSTITUTE OF TECHNOLOGY Physics Department Physics 8.07: Electromagnetism II September 15, 2012 Prof. Alan Guth PROBLEM SET 2 MASSACHUSETTS INSTITUTE OF TECHNOLOGY Physics Depatment Physics 8.07: Electomagnetism II Septembe 5, 202 Pof. Alan Guth PROBLEM SET 2 DUE DATE: Monday, Septembe 24, 202. Eithe hand it in at the lectue,

More information

. Using our polar coordinate conversions, we could write a

. Using our polar coordinate conversions, we could write a 504 Chapte 8 Section 8.4.5 Dot Poduct Now that we can add, sutact, and scale vectos, you might e wondeing whethe we can multiply vectos. It tuns out thee ae two diffeent ways to multiply vectos, one which

More information

The Omega Function -Not for Circulation-

The Omega Function -Not for Circulation- The Omega Function -Not fo Ciculation- A. Cascon, C. Keating and W. F. Shadwick The Finance Development Cente London, England July 23 2002, Final Revision 11 Mach 2003 1 Intoduction In this pape we intoduce

More information

Chapter 6 Balanced Incomplete Block Design (BIBD)

Chapter 6 Balanced Incomplete Block Design (BIBD) Chapte 6 Balanced Incomplete Bloc Design (BIBD) The designs lie CRD and RBD ae the complete bloc designs We now discuss the balanced incomplete bloc design (BIBD) and the patially balanced incomplete bloc

More information

SPE p p. Abstract

SPE p p. Abstract SPE 169333 Inflw Pefmance Relatinship Euatin f Slutin Gas-Dive Resevis unde Gavity Segegatin Rbet Padilla-Sixt, SPE, Cnsultant, and Rdlf Camach-Velázuez, SPE, Pemex E&P Cpyight 14, Sciety f Petleum Enginees

More information

COMP 551 Applied Machine Learning Lecture 4: Linear classification

COMP 551 Applied Machine Learning Lecture 4: Linear classification COMP 551 Applied Machine Learning Lecture 4: Linear classificatin Instructr: Jelle Pineau (jpineau@cs.mcgill.ca) Class web page: www.cs.mcgill.ca/~jpineau/cmp551 Unless therwise nted, all material psted

More information

20th Century Atomic Theory - Hydrogen Atom

20th Century Atomic Theory - Hydrogen Atom 0th Centuy Atomic Theoy - Hydogen Atom Ruthefod s scatteing expeiments (Section.5, pp. 53-55) in 1910 led to a nuclea model of the atom whee all the positive chage and most of the mass wee concentated

More information

Decomposing portfolio risk using Monte Carlo estimators

Decomposing portfolio risk using Monte Carlo estimators White Pape Decomposing potolio isk using Monte Calo estimatos D. Mossessian, dmossessian@actset.com V. Vieli, vvieli@actset.com Decomposing potolio isk using Monte Calo estimatos Contents Intoduction............................................................

More information

Math 1525 Excel Lab 3 Exponential and Logarithmic Functions Spring, 2001

Math 1525 Excel Lab 3 Exponential and Logarithmic Functions Spring, 2001 Math 1525 Excel Lab 3 Exponential and Logaithmic Functions 1 Math 1525 Excel Lab 3 Exponential and Logaithmic Functions Sping, 21 Goal: The goals of Lab 3 ae to illustate exponential, logaithmic, and logistic

More information

Semi-Lagrangian simulations of the diocotron instability

Semi-Lagrangian simulations of the diocotron instability Semi-Lagangian simulations of the diocoton instability Eic Madaule, Seve Adian Histoaga, Michel Mehenbege, Jéôme Péti To cite this vesion: Eic Madaule, Seve Adian Histoaga, Michel Mehenbege, Jéôme Péti.

More information

Physics 2010 Motion with Constant Acceleration Experiment 1

Physics 2010 Motion with Constant Acceleration Experiment 1 . Physics 00 Mtin with Cnstant Acceleratin Experiment In this lab, we will study the mtin f a glider as it accelerates dwnhill n a tilted air track. The glider is supprted ver the air track by a cushin

More information

Lecture 7 Topic 5: Multiple Comparisons (means separation)

Lecture 7 Topic 5: Multiple Comparisons (means separation) Lectue 7 Topic 5: Multiple Compaisons (means sepaation) ANOVA: H 0 : µ 1 = µ =... = µ t H 1 : The mean of at least one teatment goup is diffeent If thee ae moe than two teatments in the expeiment, futhe

More information

CHAPTER 4 DIAGNOSTICS FOR INFLUENTIAL OBSERVATIONS

CHAPTER 4 DIAGNOSTICS FOR INFLUENTIAL OBSERVATIONS CHAPTER 4 DIAGNOSTICS FOR INFLUENTIAL OBSERVATIONS 1 Influential bservatins are bservatins whse presence in the data can have a distrting effect n the parameter estimates and pssibly the entire analysis,

More information

PROBLEM SET #1 SOLUTIONS by Robert A. DiStasio Jr.

PROBLEM SET #1 SOLUTIONS by Robert A. DiStasio Jr. POBLM S # SOLUIONS by obet A. DiStasio J. Q. he Bon-Oppenheime appoximation is the standad way of appoximating the gound state of a molecula system. Wite down the conditions that detemine the tonic and

More information

, which yields. where z1. and z2

, which yields. where z1. and z2 The Gaussian r Nrmal PDF, Page 1 The Gaussian r Nrmal Prbability Density Functin Authr: Jhn M Cimbala, Penn State University Latest revisin: 11 September 13 The Gaussian r Nrmal Prbability Density Functin

More information

Jurnal Teknologi DIELECTRIC MEASUREMENTS FOR LOW-LOSS MATERIALS USING TRANSMISSION PHASE-SHIFT METHOD. Full Paper

Jurnal Teknologi DIELECTRIC MEASUREMENTS FOR LOW-LOSS MATERIALS USING TRANSMISSION PHASE-SHIFT METHOD. Full Paper Junal Teknlgi DIELECTRIC MEAUREMENT FOR LOW-LO MATERIAL UING TRANMIION PHAE-HIFT METHOD K. Y. Yu a *, Y.. Lee b, L. Zahid b, M. F. A. Malek c, K. Y. Lee d, E. M. Cheng e, N. H. H. Khamis a a Cmmunicatin

More information