/ / / / /DIC Bawa / Bawa DIC / CRM

Size: px

Start display at page:

Download "/ / / / /DIC Bawa / Bawa DIC / CRM"

あゆみみうら
7 years ago
Views:

2 / / / / /DIC Bawa / Bawa DIC / CRM

4 DIC DIC Bawa DIC DIC / / DIC / Z B

5 (Assael 992, 989Dck and Basu 994,Peter and Olson 999) (2002) / / / /Gvon(984), Lattn and McAlster(985),Bawa(990), Trved, Bass and Inmann(994) Gvon(984), Lattn and McAlster (985) Trved, Bass and Inmann (994)Lattn and McAlster (985)/ Bawa(990) / /

Bass and Inmann(994) Gvon(984), Lattn and McAlster (985)

6 / (2005)2004 Inman(200) / Bawa(990) / / / CRM / Bawa 2

7 2 () Guadagn and Lttle(983)

30 2000 2 2000.7.2000.2.3 2000 3 200..200.5.

8 () Guadagn and Lttle(983) t j U U jt v jt jt jt v jt jt () t C j Pjt jt u u t Pr( ) l Ct, l j (2) jt lt t j P (3) t jt j yjt yjt 0 X jt P jt e v jt e l Ct v lt (3) 4

9 v jt X jt v jt v2jt (4) / 0% / Hoer and Rdgway 984,Lattn 987, McAlster and Pessemer 982, Raju 98,984, Van Trjp,Hoyer,and Inman 996 Bawa(990)// ()/ a. Bawa(990) r jt r 2 jt s t r jt y j (5) s AABBCCC s=5 t=8 r r C8 3 r k 8 0 ( k C) j8 5

10 b. Z t r jt r 2 jt Z t t exp a exp t a (6) Z a / Bawa Z a 0 a 5 a 20 a 25 2 / (2) 3 a. x j3t j P j t j p jt x p jt j3t (7) Pj (2004) 6

11 b. x 4 0 j t x j 4 t 0 tj tj c. x 5 t j 0 j t x j 5 t 0 tj tj (3) Bawa(990)/ // / 7

12 / j r jt t j / v jt v r r 2 jt jt 2 jt / Bawa(990) 2 // / 2 / Bawa(990) 2 / 2 0 Zero-order / 3-0 0, 2 0 0, 2 0 0, 2 0 Inerta AAAAAAAAAA A 3-0, 2 0 0, 2 0 0, 2 0 VS (Varety seekng) 8

13 ACFDEADA 3-0, 2 0 Hybrd Inerta VS AAABBBCCC 3-3-, / / / 9

14 Bawa(990) Z t t exp a exp t a r jt r 2 jt a Z 3-2 Z Z 0 Z 0 Bawa(990) / LCA(Latent Class Analyss) s s s j s P j t j j, t s s P t p t S s p ( j, ) ( j ) (8), L, S,, L, S, s s 0, s, LS s S s t s 0

15 p ( s) S f f s ( y ( ) (9) y s s ) s s f T s y ( y ) p ( j ) s t s t jc t y tj t j tj (0) s s EM s s y py p p y p y p y p y () p py y py ly y y ly p p (2) p

16 Ross, Allenby and McCulloch(2005) X t j tj P j, t X tj Y tj tj Y MNL P t j X, (3) tj nd nd N( nd, nd ) (4) nd 2 nd K= nd Multnomal K ( pvec) (5) pvec Drchelt( ) (6) nd nd k N, k a (7) IW v V (8) k, v 20 V 20I J nd (J:), L, ' a , L,5 ) 0 0 J ( K 2

17 MCMC MCMC(Mrakov Chan Monte Carlo) MCMC MCMC Gbbs Samplng Metropols-Hastngs(MH) Gbbs Samplng Gbbs Samplng MH Hybrd Mrakov Chan Monte Carlo Gbbs Samplng Gbbs Samplng p (, y), j, L p j j, Gbbs Samplng (0) (0) L,, p =,2, ( ) ( ( ) ( j ( ) p ) ) (), ( ) 2, L,, L, ( ) j, L, ( ) p M, M ( ) j ( ) p,, y, y ( ) p, y ( ) N, N, N 3

Samplng p (, y), j, L p j j, Gbbs Samplng (0) (0) L,, p =,2, ( ) ( ( ) ( j ( ) p

18 Metropols-Hastngs(MH) Gbbs Samplng p (, y), j, L p j j, Metropols-Hastngs(MH) Gbbs Samplng ( y) ( ) MH (0) (0) L,, p =,2, (a) ( ) ( ) = +d ( ) d D d N0, kv k (b) ( ) ( ), y q, y y mn, ( ) ( yq ), y (9) ( ) ( ) ( yq, y 0 ), y (c)uunf(0,) () ( ) u, y ( ) ( ) ( ) N, N, N ( ) y y y ( ) y 4

19 AIC=2()+2() (Akake(973)) BIC=2()log(n)() n (Schwartz(978)) CAIC=2()(log(n)+)() AIC (Bozdogan(987)) DIC =Dbar+(Dbar-Dhat) BIC AIC AIC DbarMCMC (-2) Dhat(-2) (Spegelhalter et al(2002)) Bawa(990) 4 AIC,BIC,CAIC DIC 5

20 MCMC burn-n 5000 Geweke(992)Geweke ( r ) s ( r ) R,2,,, ( r ) r L R gs g gs g geweke s 0.R, s 0.5R ( A) s r 2 2 s 2 rs 2 s s2 s s R R R Z S g 2 R( R ) g (0) s g s 2 s S (0) g s 2 N 0, ( A) 2 w 0S g (0), M 0.3 R M Danellg( r) Z z / 2 Z z / 2 6

21 (2004) Bawa(990) 3 Bawa(990) /Bawa(990) (2002)VS InertaBawa(990) Hybrd 7

22 4-4- AIC BIC CAIC (2004) (2004) 4 /Bawa(990) DIC 4-2 DIC 8

23 4-2 DIC DIC DIC 4-3 DIC DIC 5425 DIC DIC 9

24 4-4 (j-:c-:r:r2:) VS Hybrd Hybrd Inerta Inerta 3 DIC 20

25 2 3 2

26 Bawa(990)/ Bawa / Bawa DIC 5- Bawa DIC DIC Bawa C Bawa C 0.77 Bawa Bawa C DIC 5759Bawa / 22

27 Bawa Bawa DIC // 5-3 DIC Bawa AC DIC Bawa

28 Bawa Bawa /5 30 /DIC B B Hybrd J,K 2 x KKKKJJJJKKKK 5- B 5-2 B 24

29 Hybrd Hybrd / / 5-4 / Inerta 3.78 Inerta 32.6Inerta VS VS 7.75 VS 2.33 VS Hybrd Hybrd 20.6 Hybrd 36.4 Zero-order 59.69(Inerta :3.78%;VS :7.75%;Hybrd :20.6%) 7.3 (Inerta :32.60%;VS :2.33%;Hybrd :36.43%) / / 68.5 / / / 25

30 5-5 / Zero-order 3.79 Inerta.55VS.05Hybrd.4 Zero-order Inerta VS Hybrd VS VS Inerta Hybrd 4.62 Zero-order 3.6 Zero-order VS Inerta Hybrd Bawa DIC 26

31 5-6 DIC DIC a 5 a 20 / / 5-7 / Zero-order Zero-order 40.3% Zero-order 28.7% / / / 27

32 /CRM AICBICCAIC / Bawa(990)/ / DIC Bawa(990)/ / 5 / 30 /Inerta VS Zero-order Hybrd Zero-order / 2/3 / / Zero-order Inerta VS Hybrd VS VS 28

33 / Zero-order VS Inerta Hybrd 20 / 35 /Zero-order / / / 29

34 [](989) [2](2002) Vol.202, pp.227. [3](2005), Vol.52, No4, pp3552. [4](2004) pp [5] (2005), --,. [6] Akake, H(987), Informaton theory and an extenton of the maxmum lkelhood prncple, 2 nd Internatonal Sympsum on Informaton, pp [7] Assael, H.(992), Consumer Behavor and Marketng Acton, PWS-KENT Publshng Company. [8] Bawa, K.(990), Modelng Inerta and Varety Seekng Tendences n Brand Choce Behavor, Marketng Scence, Vol.9, No.3, pp [9] Bozdogan, H. (987), Model-selecton and Akake's Informaton Crteron (AIC): The general theory and ts analytcal extensons, Psychometrka, Vol52, pp [0] Dck, A. S. and Basu, K.(994), Customer Loyalty: Toward an Integrated Conceptual Framework, Journal of the Academy of Marketng Scence, Vol.22, pp.993. [] Geweke, J.(992), "Evaluatng the Accuracy of Samplng-Based Approaches to the Calculaton of Posteror Moments, J. M. Bernardo, J. O. Berger, A. P. Dawd and A. F. M. Smth, eds., Bayesan Statstcs, 4, pp [2] Gvon, M.(984), Varety Seekng through Brand Swtchng, Marketng Scence, Vol.3, No., pp.22. [3] Guadagn, P. M. and Lttle, J. D.C. (983), A Logt Model of Brand Choce calbrated on Scanner Data, Marketng Scence, Vol.2, No.3, pp

35 [4] Hoyer, W. D. and Rdgway, N. M. (984), Varety Seekng as an Explanaton for Exploratory Purchase Behavor: A Theoretcal Model, Advances n Consumer Research, Vol., pp49. [5] Inman, J. J.(200), The role of sensory-specfc satety n attrbute-level varety seekng, Journal of Consumer research, Vol.8,pp [6] Lattn, J. M.(987), A model of Balanced Choce Behavor, Marketng Scence, Vol.6, pp4865. [7] Lattn, J. M. and McAlster, L.(985), Usng a Varety-Seekng Model to Identfy Substtute and Complementary Relatonshps Among Competng Products, Journal of Consumer research,vol.22,pp [8] McAlster, L. and Pessemer, E.(982), Varety seekng Behavor: An Interdscplnary Revew, Journal of Consumer research,vol.9, pp [9] Peter, J. P. and Olson, J. C. (999), Consumer Behavor and Marketng Strategy, Mcgraw-Hll/Irwn Seres n Marketng. [20] Raju, P.S. (98), Theores of Exploratory Behavor:Revew and Consumer Research Implcatons, Research n Marketng, Vol.4, pp [2] Raju, P.S. (984), Exploratory Brand Swtchng: An Emprcal Examnaton of Its Determnants, Journal of Economc Psychology, Vol.5, pp [22] Ross, P. E., Allenby, G. M. and McCulloch, R. (2005), Bayesan Statstcs and Marketng, John Wley and Sons. [23] Schwarz, G. (978), Estmatng the Dmenson of a Model, The Annals of Statstcs, Vol.6, No.2, pp [24] Spegelhalter, D. J., Best, N. G., Carln, B. P. and van der Lnde, A. (2002), Bayesan Measures of Model Complexty and Ft, Journal of the Royal Statstcal Socety Seres B, pp [25] Trved, M., Bass F. M. and Rao, R. C. (994), A model of stochastc varety seekng, Marketng Scence, Vol.3, No.3, pp [26] Van Trjp, H. C. M., Hoyer, W. D. and Inman, J. J. (996), Why Swtch? Product Category-Level Explanatons for True Varety-Seekng Behavor, Journal of Consumer research, Vol.33, pp

36 32

37 A B 33

38 C Hybrd 34

39 D DIC E / DIC Bawa / a /a 25 35

JGSS統計分析セミナー2009－傾向スコアを用いた因果分析－

JGSS統計分析セミナー2009－傾向スコアを用いた因果分析－日本版総合的社会調査共同研究拠点研究論文集 [10] JGSS Research Seres No.7 JGSS 2009 JGSS JGSS Statstcal Analyss Semnar: Causalty Analyss based on the Propensty Score Kana MIWA JGSS Research Center Osaka Unversty of Commerce