A COMPARISON OF ADJUSTED BAYES ESTIMATORS OF AN ENSEMBLE OF SMALL AREA PARAMETERS

Size: px
Start display at page:

Download "A COMPARISON OF ADJUSTED BAYES ESTIMATORS OF AN ENSEMBLE OF SMALL AREA PARAMETERS"

Transcription

1 STATISTICA, anno LXIX, n. 4, 009 A COMPARISON OF ADJUSTED BAYES ESTIMATORS OF AN ENSEMBLE OF SMALL AREA PARAMETERS Enrco Fabrz 1. INTRODUCTION In recen years saple surveys have been characerzed by a growng deand for esaes of populaon descrpve quanes for doans (or 'areas') obaned classfyng he arge populaon accordng o geography or oher crera. As he saple poron peranng o doans s ofen oo sall o allow for relable esaon usng sandard desgn-based esaors, sall area esaon ehods have becoe a relevan research opc (see Rao, 003 or Lahr and Jang, 006 for a general nroducon). Eprcal and herarchcal Bayes ehods are an poran chaper of sall area esaon heory and are also wdely appled n pracce (see Rao, 003 Chapers 9 and 10 and he references heren). The basc dea behnd hese ehods s o rea doan descrpve quanes of neres (e.g. eans, oals, proporons) as rando and o esae he usng soe suary of her poseror dsrbuon, ypcally he poseror ean, ofen referred o as 'Bayes esaor' (Ghosh, 199). Bayes esaors ay be very effecve n provng he precson (saplng Mean Square Error) of 'drec' desgn-unbased (or desgn-conssen) esaors, bu hs proveen s ofen acheved a he cos of shrnkng he esaes oward a synhec esaor whch s obaned poolng ogeher daa fro all areas under sudy. For hs reason, Bayes esaors ay be proven o be poor for esang he acual dsrbuon funcon of a populaon ('enseble') of sall area paraeers (Lous, 1984, Heady and Ralphs, 004). The neres n he dsrbuon funcon ay be crucal when sall area esaes are used n subsanve applcaons such as he analyss of regonal dspares n he dsrbuon of econoc ndcaors of povery and ncoe nequaly (Fabrz e al. 005). A proper represenaon of he dsrbuon of he enseble of he paraeers s, for hese purposes, as poran as o dspose of relable esaes for each sub-naonal regon. In hs paper we dscuss he popular Fay-Herro odel (Fay and Herro, 1979) and a se of adjused esaors assocaed o. Wh adjused esaors we ean esaors of he sall area paraeers ha enjoy accepable properes wh respec o he esaon of eprcal dsrbuon funcon (EDF) or oher nonlnear funconals of he populaon ('enseble') of sall area paraeers.

2 70 E. Fabrz The an goal of he paper s o revew adjused esaors whn he fraework of herarchcal Bayesan odelng and o copare her frequens properes by eans of a Mone Carlo exercse. In parcular we focus on: ) he dsance beween he esaed and he rue dsrbuon funcon; ) effcency as easured by ean square error; ) robusness wh respec o he falure of he assupon of noraly of he rando effecs. We ephasze frequens properes snce hese are usually relevan for praconers and ore falar o fnal daa users. Moreover, we wll use he sae sulaon exercse o evaluae wheher poseror ean square errors, a naural easure of uncerany assocaed o adjused Bayes esaors, are also good frequens easures of varably. As ancpaed, we consder he effecs of sspecfcaon concernng he rando effecs: n parcular we focus on wrong dsrbuonal assupons. Ths ype of sspecfcaon s que lkely n pracce snce rando effecs are no drecly observable and deparures fro noraly are dffcul o deec (see Snharay and Sern, 003). We explore he properes of he adjused esaors when he odels assue noraly bu he rando effecs are acually generaed by an alernave dsrbuon. The paper s organzed as follows. In secon, we shorly dscuss he falure of Bayes esaors assocaed o he unvarae Fay-Herro odel as esaors of he varance (and hus of he EDF) of he 'enseble' of paraeers. Aong he any adjused esaors dscussed n he leraure we focus on consraned Bayes esaors (Ghosh, 199), consraned lnear Bayes esaors (Spjøvoll and Thosen, 1987) and a sulaneous esaon ehod proposed by Zhang (003). These esaors are revewed n secon 3. The sulaon exercse and he ools used n coparsons are nroduced n secon 4. Alhough all sulaons use populaons generaed under noraly, we wll focus on he accuracy of EDF esaon and no jus on ean and varance (as n Judkns and Lu, 000) snce wh a fne nuber of areas, he EDF of he populaon of area paraeers ay show soe slgh devaon fro noraly. Secon 5 focuses on sulaon s resuls; he behavour of adjused esaors when he noraly of rando effecs fals s dscussed n Secon 6. Concludng rearks are provded n Secon 7.. FAILURE OF BAYES ESTIMATORS AS ENSEMBLE ESTIMATORS IN THE FAY-HERRIOT MODEL The Fay-Herro odel ay be descrbed by he followng se of assupons: y e (1) v nd x () e N(0, ) (3)

3 A coparson of adjused Bayes Esaors of an enseble of sall area paraeers 71 nd v v N(0, ) (4) where { y } 1 s a collecon of 'drec' desgn-unbased (or approxaely desgn-unbased) esaors of a se of sall area populaon paraeers { }; { } s he se of assued known desgn-based varances assocaed o drec esaors and x a k 1 vecor of auxlary nforaon accuraely known for area. Moreover s assued ha Eev ( ) 0. (5) Sall area analyses are soewha dosyncrac as hey x randozaon and odel based probably spaces. More precsely, once denoed E D( ), V D( ) he expecaon and varance wh respec o he randozaon (desgn) dsrbuon and E M (), V M () he oens relaed o he odel or daa generang process, assupons (1) (4) ply ha EM( y ) ED( y) and VM( y ) VD( y), ha s he frs wo oens of y accordng o he odel (condonal on ) and randozaon dsrbuons are he sae. To be conssen n noaon le's also wre EM( ) x, VM( ) v and EM( ev) 0. Assung and v as known, he poseror eans of under odel (1) (5) are gven by ˆ B 1 y (1 ) x, where v ( v ) s a shrnkage facor ha gves o he drec esaor a wegh ha s a decreasng funcon of s saplng varance. We have ha drec esaors { y } are overdspersed wh respec o he underlyng { } and he sall area esaors are under-dspersed. The followng proposon holds. Proposon 1 Under odel (1) (5) and assung 1) we have ha: 1 EM ( ) v xx 1 (6) 1 1 wh xx ( 1) ( x x)( x x ), x x, ) 3) 1 v EMED ( y y) xx v xx 1 1 (7) 1 ˆ ˆ EMED ( ) v xx v xx 1 1 (8)

4 7 E. Fabrz A proof ay be found n he appendx. Noe ha does no represen a resrcve condon bu s helpful o oban sple forulas. We ay observe ha he shrnkage facor rules boh overdsperson of drec esaes and underdsperson of poseror eans. 3. ADJUSTED BAYES ESTIMATORS We consder hree dfferen adjused Bayes esaors assocaed o he Fay- Herro odel : ) he consraned Bayes, ) he consraned lnear Bayes (Spjøvoll and Thosen ehod), ) he esaors based on a sulaneous esaon ehod proposed by Zhang (003). These esaors wll be revewed whn a herarchcal Bayes fraework; ha s, we do no assue he hyperparaeers, v as known bu specfy a pror dsrbuon for he. In wha follows we denoe he daa on whch he analyss s condoned as z{ y,, x } 1. We ancpae ha all adjused esaors ˆ are suares of he poseror dsrbuons p( z ) dfferen fro he poseror ean ˆ HB E( z ) and are herefore subopal wh respec o quadrac loss. Uncerany assocaed o hese alernave poseror suares ay be easured by he poseror ean square error: PMSE( ˆ ) ( z ) ( ˆ ˆ ) (9) * * HB V We denoe E ( ) he expecaon wh respec o he odel ong he subscrp M, snce randozaon oens are no longer nvolved. Of course, PMSE( ) V( z ); he beer represenaon of he Eprcal Dsrbuon Funcon of he populaon of Sall Area paraeers s pad a he prce of soe loss of effcency. 3.1 The Consraned herarchcal Bayes esaor Consraned Bayes esaors have been nroduced and dscussed by Lous (1984) under noraly and by Ghosh (199) under less resrcve dsrbuonal assupons. The a s o oban a se of esaors { }, 1,...,, opal under quadrac loss and sasfyng he followng consrans: 1) ˆ HB 1 ) ( 1) ( ) E ( ) z 1 1 The consraned herarchcal Bayes (CHB) esaors (see Rao, 003, secon 10.13) are gven by:

5 A coparson of adjused Bayes Esaors of an enseble of sall area paraeers 73 ˆCHB ˆ HB ˆHB ˆ HB a( z )( ) (10) where V( 1 ) z az ( ) 1 ( 1) 1 ˆ ˆ 1 ( HB HB ) The poseror ean square error ( ˆCHB PMSE ) ay be used o evaluae he uncerany assocaed o hs esaor. 3. The consraned herarchcal lnear Bayes Esaor Le's suppose, for he oen, ha he hyperparaeers, v are known. The Consraned Lnear Bayes esaor of s a suary of he poseror dsrbuon n he for ˆL a y b sasfyng he consrans: 1) ( ˆL E ) x, 1 ) E( ˆ x ). L v Noe ha when he poseror ean ay be expressed n lnear for, he dsrbuon enjoys poseror lneary (Goldsen, 1975), a condon ha holds for a varey of conjugae odels. The consraned lnear Bayes esaor s gven by: ˆ CLB 1/ 1/ y (1 ) x (11) 1 wh v ( v ) (see Spjøvoll and Thosen, 1987 and Rao, 003, secon 9.8). Ths esaor owes s populary o s slary o ˆB : s sll a lnear cobnaon of y and x ha, wh respec o ˆB, pus ore wegh on he 'drec' esaor y, whereby leadng o a se of esaes less shrunken oward he synhec coponen. The esaor (11) ay be hough as condonal on (, v ). We sply propose o defne he Consraned Herarchcal Lnear Bayes (CHLB) esaor as ˆ CHLB E v (, ) ( ˆCLB ) (1) z where E (, v z s he expecaon aken wh respec o he poseror dsrbuon ) of, v. An explc forula for (1) depends on he chosen pror dsrbuons

6 74 E. Fabrz and ay be n general dffcul o work ou. Noneheless (1) ay be easly approxaed usng he oupu of Markov Chans Mone Carlo (MCMC) algorhs of coon use for he analyss of herarchcal odels. 3.3 The sulaneous esaon ehod proposed by Zhang Gven he se { } of he area paraeers of neres, le { ( )} be he assocaed ordered se ( (1) ()... ( )). Then E( ( ) z ) s he bes predcor of under quadrac loss and { } s he bes 'enseble' esaor of { } under quadrac loss. The se of esaors { } s no area specfc n ha s sngle eleens are no assocaed o specfc areas. To ach he { } wh he sall areas Zhang (003) proposes, under an eprcal Bayes esaon approach, o esae he ranks of { } usng hose of he { E( z )} se. By he way, he ranks of he poseror eans ay be poor esaors of acual ranks, especally f here s consderable varably n he poseror varances. Followng Ghosh and Ma (1999) we propose rˆ E( rank( z )), he poseror expecaon of ranks, as he esaor of ranks requred n order o ach he enseble esaor { } wh he areas. In he conex of herarchcal Bayes odelng, hs esaor of ranks ay be easly approxaed fro he oupu of MCMC algorhs. More specfcally, we can rank he ( s ) z fro any draw s of he Markov Chan afer convergence. Then we can approxae r ˆ averagng he ranks rank( ( s ) z ) MC 1 S over all draws, obanng rˆ S rank( ( ) ) s 1 s z where S s he nuber of eraons of Markov Chan afer convergence used for he esaon of he poseror dsrbuon. To suarze, he esaor based on Zhang deas pleened n he conex of herarchcal Bayes odelng s gven by: ˆ ZHB (13) rˆ wh r ˆ approxaed by r ˆMC when he poseror dsrbuons are obaned usng MCMC algorhs. 4. A SIMULATION EXPERIMENT UNDER THE ASSUMPTION OF NORMALITY FOR THE RAN- DOM EFFECTS In hs secon we nroduce a sulaon experen whose a s o copare he effecveness of he adjused esaors dscussed above n correcng he overshrnkage. We copare also her effcency and consder wheher he PMSE are adequae esaors of her frequens ean square errors. More

7 A coparson of adjused Bayes Esaors of an enseble of sall area paraeers 75 specfcally, we copare he drec esaors ˆ { y }, he poseror eans ˆHB { ˆHB } and he varous adjused esaors ˆCHB { ˆCHB }, ˆCHLB { ˆCHLB } and ˆ ZHB { ˆ ZHB }. The sulaon s based on R=1,000 Mone Carlo (MC) saples, and all coparsons are referred o he eprcal dsrbuon of he varous esaors n hs replcaon space. Daa are generaed accordng o he Fay-Herro odel (1) (5) seng for splcy x 0. We consder boh he case of oderae and large nuber of areas seng 30, 100. We se v 1 and consder hree dfferen confguraons of desgn varances. They are chosen n he followng way: we dvde he se of areas n fve groups. Varances vary across groups bu are consan whn he. All confguraons are llusraed n Table 1. They dffer accordng o nforaveness of 1 drec esaors, as easured by v ( v ). We evaluae he nforaveness of drec esaors n coparson o he dsperson of he values around he synhec coponen. Populaon 1 descrbes a suaon where drec esaors show a wde range of nforaveness ( [0.11,0.91] ); Populaon descrbes a suaon n whch drec esaors are poorly nforave ( [0.11, 0.5] ), whle n Populaon 3 we sudy he case of raher srongly nforave drec esaors ( [0.5, 0.91] ). DIR TABLE 1 Dfferen confguraons of desgn varances for he sulaon Desgn varances ,..., 1,..., 1,..., 1,..., 1,..., Populaon Populaon Populaon We do no consder equal saplng varances,.e. snce hs s seldo he case n pracce; oreover ay be proven ha, for x and, v known, ˆ CHB ˆ CHLB for (Rao, 003, secon 9.6). As a consequence, even f we do no assue and v as known and work wh a fne nuber of areas, we ay expec he wo esaors o show close perforances. When odelng, s assued ha and v are unknown wh pror dsrbuons p(, v ) p( ) p( v ) wh N(0, K), v Unf (0, L ) where K 100 and L 0 are large consans wh respec o he scale of he daa.

8 76 E. Fabrz Ths prors guaranee properness of he poseror dsrbuons, very ld pac on poseror dsrbuons and good behavor (fas convergence and good xng) of MCMC algorhs (Gelan, 006). To copare esaors, we consder he ndcaors descrbed below. ) Overshrnkage correcon. Le's defne R * 1 r r 1 AV( ˆ ) R v ( ) (14) ˆ ˆ ˆ, and * { DIR, HB, CHB, CHLB, ZHB}., we expec hs ndcaor o be larger han 1 for, less han 1 for ˆHB and close o 1 for he reanng esaors. ) Kologorov-Srnov dsance. For each eraon r, he Kologorov-Srnov dsance beween he EDF of he adjused esaor ˆ * { ˆ r, r} and ha of he rue values { r, } denoed respecvely as F * and F T s calculaed as D (, ) ax F ( u ) F ( u ) where j 1,..., and u (, ) s he * 1 * * where v ( r ) ( 1) ( 1, r r ) Snce we se x and v 1 ˆDIR r r r j * j, r T j, r r r r vecor obaned poolng ogeher he rue values and hose obaned wh he adjused predcor for he r-h MC replcaon. The dsances calculaed a each eraon are hen averaged over MC replcaons o oban 1 R D(, ) R D (, ). For he ease of coparson we repor r 1 ˆ r r r * D(, ) (, ) (15) HB D D( ˆ, ) whereby assung he non adjused herarchcal Bayes esaors as a benchark. ) Anderson-Darlng dsance. Anderson and Darlng (1954) nroduced a goodness-of-f sasc ha can be used o evaluae he dsance of an EDF fro a connuous reference dsrbuon. Copared o he Kologorov-Srnov dsance, s known o be nfluenced o a greaer exen by he dscrepances n he als of he dsrbuon. In our parcular conex he 'eprcal' Anderson-Darlng dsance s defned as follows 1 F ( u ) F ( u ) A (, ) ( F ( u )) * jr, T jr, r 1 (0,1) T j, r 4 j 1 FT( u j, r )[1 FT( u j, r )] The copued dsances are hen averaged over all R replcaons, and ˆ * A (, ) (, ) (16) HB A s repored n resuls. A ( ˆ, )

9 A coparson of adjused Bayes Esaors of an enseble of sall area paraeers 77 v) Frequens effcency. We evaluae he average pac of he adjusen on he uncondonal frequens MSE of sall area predcors defned as MSE( ) E( ) (see Rao, 003, secon 6.). We esae he MSE( ) * 1 * by eans of s Mone Carlo approxaon ˆ R se ( ) ( ˆ MC R 1 r, r, ). r These quanes are area-specfc. We jus focus on he ean of her dsrbuon across areas: ase ( ˆ ) se ( ˆ ) (17) * 1 * MC 1 MC v) Frequens evaluaon of PMSE. Frequens properes of (9) are evaluaed usng he followng easure of relave bas: apse( ˆ ) ( ˆ ) (18) 1 R * * 1 R pse 1 MC r r, 1 semc ( ) * where pse ( ˆ MC r, ) s he pse ( MC ) calculaed usng daa fro he r-h draw of he Mone Carlo exercse. Moreover noe ha ( ˆHB psemc ) error reduces o poseror varance. Codes are wren n R (R Developen Core Tea, 006). For MCMC calculaons we used he Brugs package (Thoas and O'Hara, 006) whch recursvely calls he MCMC dedcaed sofware OpenBUGS (Thoas e al. 006). As for echncal deals concernng MCMC calculaons, we generae saples of sze 0,000 for all chans deleng a conservave 'burn n' saple of sze 5,000. In fac, he relavely sple noral odels eployed n he sulaons have all shown very fas convergence raes. Convergence has been checked by eans of sandard convergence sascs (Cowles and Carln, 1996). 5. SIMULATION RESULTS For brevy, we repor resuls for he case 100 and dscuss he case 30 only when dfferences are rearkable or he coparson hghlghs poran pons. Resuls abou he shrnkage correcon are dsplayed n able. I s apparen ha drec esaes are overdspersed, wh overdsperson ncreasng wh he average varance of drec esaors; herarchcal Bayes esaors are underdspersed n all suaons, and ore serously so when he drec esaes convey lle nforaon (Populaon ). More poran, all adjused esaors approxaely elnae he overshrnkage. Flucuaons of values around 1 do no see o follow any sgnfcan paern. The correcon of overshrnkage does no see o be nfluenced by he nuber of areas.

10 78 E. Fabrz TABLE Overshrnkage correcon as easured by he ndcaor AV( ), 100 ˆDIR ˆHB ˆCHB ˆCLHB ˆZHB Populaon Populaon Populaon Table 3 shows resuls based on he Kologorov-Srnov dsance. All adjused esaors clearly prove he perforances of ˆHB and he proveen s larger when 100 wh respec o he case of 30. Aong adjused esaors, ˆZHB eerges clearly as he bes. Noe ha, gven he sze of Mone Carlo (MC) errors, all observed dfferences appearng n able 3 can be aken as 95% sgnfcan. Moreover, noe ha he advanage of ˆZHB over he oher adjused ehods s less pronounced for Populaon (poorly nforave drec esaes). Ths akes sense snce he esaon of ndvdual ( ) requres ore nforaon han he global adjusen on whch ˆCHB and ˆCLHB are based. ˆCHB and ˆCLHB perfor closely and none of he wo sees preferable. TABLE 3 Rao of he Kologorov-Srnov dsances beween esaed and acual EDF averaged over MC replcaons dvded by he sae quany calculaed for he ˆHB esaors,.e. D (, ), 100 Populaon ˆDIR ˆHB ˆCHB ˆCLHB ˆZHB Populaon Populaon Populaon Table 4 presens he resuls relaed o he Anderson-Darlng dsance. We noe ha ˆHB s n an neredae poson beween he drec esaors (ha are he wors perforers over all sengs) and he adjused esaors ha are beer. Aong adjused esaors ˆZHB s clearly beer han he oher wo excep for Populaon, characerzed by poorly nforave drec esaes where, by he way, sll perfors a lle beer. We ay hen conclude ha he esaon ehod proposed by Zhang urns ou o be he bes wh respec o boh consdered easures of dsance and gves s bes when he nuber of areas s large and he drec esaes are no oo precse. TABLE 4 Rao of he Anderson-Darlng dsances beween esaed and acual EDF averaged over MC replcaons dvded by he sae quany calculaed for he ˆHB * esaors,.e. A ( ˆ, ), 100 Populaon ˆDIR ˆHB ˆCHB ˆCLHB ˆZHB Populaon Populaon Populaon

11 A coparson of adjused Bayes Esaors of an enseble of sall area paraeers 79 The resuls relaed o he repeaed saplng effcency, as easured by he eprcal uncondonal Mean Square Error are shown n Table 5, where he adjusen of ˆHB has, as expeced, a cos n ers of effcency. The ncrease n Mean Square Errors depends on he precson of drec esaors. When he precson s hgh (Populaon 3), he rse s around 10%, bu when s low, ase MC ( ) are 0% or even 30% (n he case of ˆCLHB ) hgher han n he case of ˆHB. Noneheless we ay noe ha he proveen wh respec o he drec esaors reans subsanal. Moreover ˆCHB and ˆZHB show slar perforances, whle ˆCLHB urns ou o be a lle less effcen. TABLE 5 Effcency of he consdered esaors as easured by ase MC ( ), 100 ˆDIR ˆHB ˆCHB ˆCLHB ˆZHB Populaon Populaon Populaon A ore dealed analyss of he dsrbuon of ase MC ( ) across areas (for whch we do no show ables) hghlghs he very dfferen behavor of ˆCLHB wh respec o ˆCHB and ˆZHB : perfors clearly beer when drec esaes are ore precse han he average and far worse n he case of areas characerzed by he os precse drec esaes. Ths behavor depends on he naure of he esaor. Fro (11) we ay noe ha, wh respec o poseror ean, ˆCHB gves less wegh o he synhec coponen and ore o he drec one. When y s very precse, hs leads o ore effcen esaors han oher adjused ehods; unforunaely y receves ore wegh even when s unrelable, hus producng a large loss n effcency wh respec o and he oher adjused ehods. TABLE 6 Average rao of he esaed PMSE o he MSE : apse( ), 100 ˆHB ˆCHB ˆCLHB ˆZHB Populaon Populaon Populaon Table 6 repors an evaluaon of frequens properes of he Poseror Mean Square Error defned n (9). I s apparen ha hs Bayesan uncerany easure, represens a sensble easure of varably also wh respec o repeaed saplng; n fac n all cases s approxaely unbased; alhough should be noed ha hs propery holds on average wh respec o he se of areas beng suded. The propery, ha was known o hold for he poseror varances as frequens var-

12 80 E. Fabrz ably easures of ˆHB under careful choce of he prors (Ganesh and Lahr, 008), s hen exensble o he case of poseror ean square errors, a leas for he pror dsrbuons chosen n our sulaon exercse. 6. PROPERTIES OF THE ADJUSTED ESTIMATORS UNDER FAILURES OF THE NORMALITY OF RANDOM EFFECTS In hs secon we copare he perforance of adjused esaors already consdered when he assued noraly of rando effecs does no hold, ha s, we evaluae wheher hey are robus wh respec o hs deparure fro he assupons under whch hey are obaned. We generae populaon daa accordng o he sae sulaon experen dscussed n Secon 4, changng he dsrbuon assued for he rando effecs. We consder: 1) v Laplace(0,1/ ) and ) v Exp(1). In boh cases Vv ( ) 1, so he nerpreaon of Populaon 1, Populaon, Populaon 3 n ers of nforaveness of he drec esaors reans unchanged. Case 1) represens a ld devaon fro noraly whle case ) represens a ore serous deparure fro hs assupon. We noe ha he perforances of herarchcal Bayes esaors of he area eans rean approxaely unchanged when rando effecs are generaed by a Laplace dsrbuon bu a noral odel s assued. Ths robusness wh respec o oderae falures of he assupons on he dsrbuon of he rando effecs s noed n Snharay and Sern (003) and Fabrz and Trvsano (009). When rando effecs are generaed fro an Exponenal dsrbuon, he deeroraon of he perforances of HB s subsanal, even hough no caasrophc. Under he assupon on he drec esaors varances of Populaon 1 he Kologorov-Srnov dsance grows by a 30%, he Anderson-Darlng by a 60%, whle he Mean Square Error averaged over he se of all he areas only by 10%. In Table 7 resuls abou he Kologorov-Srnov and he Anderson- Darlng dsance are repored for 100. Resuls relaed o oher ndcaors and 30 are no repored snce hey are subsanally slar. Fro Table 7 we ay noe ha all adjused predcors provde saller proveens wh respec o HB han n he case of noraly; hs eans ha he esaon of he Eprcal Dsrbuon Funcon of he area averages or soe of s feaures s ore sensve o wrong dsrbuonal assupons han he esaon of ndvdual area averages. The perforance of he adjused predcors are now closer han under noraly of rando effecs. ˆZHB reans slghly beer n ers of Kologorov- Srnov dsance, whle ˆCHB s a lle beer f we consder he Anderson- Darlng dsance and ase MC ( ). Snce ˆZHB urned ou o be beer under noraly, we ay conclude ha s perforances deerorae ore when hs as-

13 A coparson of adjused Bayes Esaors of an enseble of sall area paraeers 81 supon fals wh respec o oher adjused esaors. Ths akes sense, snce ˆZHB akes an heaver use of he assupon of noraly n he esaon of { } han he oher predcors, whose correcons are based on he esaon of ( ) oens. Noe, however, ha he perforances of ˆZHB rean always coparable o hose of oher adjused predcors even under bg deparures fro noraly of rando effecs. TABLE 7 Coparson of adjused esaors under falures of he dsrbuonal assupon on rando effecs, 100 Acual dsrbuon of he v Populaon ˆDIR ˆHB ˆCHB ˆCLHB ˆZHB Laplace Exponenal Laplace Exponenal D (, ) Populaon Populaon Populaon Populaon Populaon Populaon A (, ) Populaon Populaon Populaon Populaon Populaon Populaon CONCLUSIONS In hs paper we dscuss and copare hree dfferen ehods for adjusng a se of sall area esaors n order o prove esaon of he EDF of he 'enseble' of sall area paraeers. Two of hese predcors (CHB and CLHB) are well known n he leraure, whle he hrd (ZHB) s ore recen. The evaluaon of he perforances of he hree esaors s based on a sulaon exercse coverng a range of suaons whch are poenally relevan for sall area praconers. A frs concluson fro hs analyss s ha all he ehods consdered ee he goal of correcng overshrnkage and leadng o ore realsc esaes of EDF of he 'enseble' of sall area eans. On he oher hand, adjused esaors are less effcen han poseror eans, bu her gan n precson wh respec o drec esaors reans subsanal. Zhang s predcor sees as beer han he oher wo when he noraly of rando effecs holds. We also consdered falures of hs assupon, generang populaons wh Laplace and Exponenally dsrbued rando effecs. In hs laer cases Zhang s predcor and he beer known CHB predcor show close perforances. Thereby Zhang s predcor s ore sensve o he falure of dsrbuonal assupons. We suded he frequens behavor of easures of uncerany assocaed o adjused herarchcal Bayes esaors, fndng ha, a leas for he pror dsrbu-

14 8 E. Fabrz ons chosen n he sulaon exercse, he poseror MSEs have good frequens properes and ay be assued by pracconers as accepable easures of frequens varably. Dpareno d Scenze Econoche e Socal Unversà Caolca del S. Cuore, sede d Pacenza ENRICO FABRIZI APPENDIX Proof of (6). Now 1 1 EM ( ) E ( x x)( x x) v v 1 as M( ) M( ) [ M( )] v xx E V E and 1 M( ) M( ) [ ( )] v E V E xx To prove (7) noe ha E 1 1 M E D ( ) M D 1 y y 1 1 E E y y 1. As D( ) D( ) [ D( )] E y V y E y and 1 D( ) D( ) [ D( )] E y V y E y we have ha 1 1 EMED y y EM ( 1) Wh he sae arguen used n he proof of (6) we oban 1 1 EM ( 1) v ( )( ) x x x x 1 Evenually, o prove (8) noe ha y (1 ) x and B

15 A coparson of adjused Bayes Esaors of an enseble of sall area paraeers 83 1 ˆB ˆ B 1 ( ) ( y y) (1 ) xx (1 ) ( y y)( x x) 1 1 As a consequence 1 ˆB ˆ B EMED ( ) 1 1 (1 ) (1 ) xx v xx xx v xx REFERENCES T.W. ANDERSON AND D.A. DARLING (1954) A es of goodness of f, Journal of he Aercan Sascal Assocaon, 49, pp M.K. COWLES, B.P. CARLIN (1996) Markov Chan Mone Carlo convergence dagnoscs: a coparave revew, Journal of he Aercan Sascal Assocaon, 91, pp E. FABRIZI, M.R. FERRANTE AND S. PACEI (005) Esaon of povery ndcaors a sub-naonal level usng ulvarae sall area odels. Sascs n Transon, 7, pp E. FABRIZI AND C. TRIVISANO (009) Robus Lnear Mxed Models for Sall Area Esaon, Journal of Sascal Plannng and Inference, 140, R. FAY R. AND R.A. HERRIOT (1979) Esaes of ncoe for sall places: an applcaon of Jaes-Sen procedures o Census daa, Journal of he Aercan Sascal Assocaon, 74, pp N. GANESH AND P. LAHIRI (008), A new class of average oen achng prors, Boerka, 95, pp A. GELMAN (006) Pror dsrbuon for varance paraeers n herarchcal odels, Bayesan Analyss, 1, pp M. GHOSH (199) Consraned Bayes esaon wh applcaons, Journal of he Aercan Sascal Assocaon, 87, M. GHOSH AND T. MAITI (1999) Adjused Bayes esaors wh applcaons o Sall Area Esaon, Sankhya, Ser. B, 61, pp H. GOLDSTEIN (1975) A Noe on soe Bayesan Nonparaerc Esaes, Annals of Sascs, 3, pp P. HEADY AND M. RALPHS (004) Soe fndngs of he EURAREA projec and her plcaons for sascal polcy, Sascs n Transon, 6, pp D.R. JUDKINS AND J. LIU (000) Correcng he Bas n he Range of a Sasc across Sall Areas, Journal of Offcal Sascs, 16, pp P. LAHIRI, J. JIANG (006) Mxed odel predcon and sall area esaon, Tes, 15, pp T.A. LOUIS (1984) Esang a Populaon of Paraeers Values Usng Bayes and Eprcal Bayes Mehods, Journal of he Aercan Sascal Assocaon, 79, pp J.N.K. RAO (003) Sall Area Esaon, John Wley and Sons, New York. R DEVELOPMENT CORE TEAM (006) R: A language and envronen for sascal copung. R Foundaon for Sascal Copung, Venna, Ausra. ISBN , URL downloadable a hp://

16 84 E. Fabrz S. SINHARAY, H.S. STERN (003) Poseror predcve odel checkng n herarchcal odels, Journal of Sascal Plannng and Inference, 111, pp E. SPJØTVOLL, I. THOMSEN (1987) Applcaon of soe eprcal Bayes ehods o Sall Area Esaon, Bullen of he Inernaonal Sascal Insue,, pp A. THOMAS, B. O'HARA (006) The BRugs package, Sofware and Docuenaon, downloadable a hp://cran.r-projec.org/org/packages/brugs. A. THOMAS, B. O HARA, U. LIGGES, S. STURZ (006) Makng BUGS Open, R News, 6, pp L.C. ZHANG (003) Sulaneous esaon of he ean of a bnary varable fro a large nuber of sall areas, Journal of Offcal Sascs, 19, pp SUMMARY A coparson of adjused Bayes Esaors of an enseble of sall area paraeers Wh enseble properes of sall area esaors, we ean her ably o reproduce he Eprcal Dsrbuon Funcon (EDF) characerzng he collecon of underlyng sall area paraeers (eans, oals). Good enseble properes ay be relevan when esaon of non-lnear funconals of he EDF of sall area paraeers (such as her varance) s needed. Sall area esaors assocaed o he popular Fay-Herro odel are consdered. Bayes esaors,.e. poseror eans, do no enjoy of good enseble properes. In hs paper hree dfferen adjused predcors are copared, by eans of a sulaon exercse, under he assupon of correcly specfed odel. As he dsrbuonal assupons on he rando effecs are dffcul o assess, he consdered predcors are copared also wh respec o her robusness o he presence of falures n he dsrbuonal assupons on he rando effecs.

THEORETICAL AUTOCORRELATIONS. ) if often denoted by γ. Note that

THEORETICAL AUTOCORRELATIONS. ) if often denoted by γ. Note that THEORETICAL AUTOCORRELATIONS Cov( y, y ) E( y E( y))( y E( y)) ρ = = Var( y) E( y E( y)) =,, L ρ = and Cov( y, y ) s ofen denoed by whle Var( y ) f ofen denoed by γ. Noe ha γ = γ and ρ = ρ and because

More information

V.Abramov - FURTHER ANALYSIS OF CONFIDENCE INTERVALS FOR LARGE CLIENT/SERVER COMPUTER NETWORKS

V.Abramov - FURTHER ANALYSIS OF CONFIDENCE INTERVALS FOR LARGE CLIENT/SERVER COMPUTER NETWORKS R&RATA # Vol.) 8, March FURTHER AALYSIS OF COFIDECE ITERVALS FOR LARGE CLIET/SERVER COMPUTER ETWORKS Vyacheslav Abramov School of Mahemacal Scences, Monash Unversy, Buldng 8, Level 4, Clayon Campus, Wellngon

More information

Department of Economics University of Toronto

Department of Economics University of Toronto Deparmen of Economcs Unversy of Torono ECO408F M.A. Economercs Lecure Noes on Heeroskedascy Heeroskedascy o Ths lecure nvolves lookng a modfcaons we need o make o deal wh he regresson model when some of

More information

Robustness Experiments with Two Variance Components

Robustness Experiments with Two Variance Components Naonal Insue of Sandards and Technology (NIST) Informaon Technology Laboraory (ITL) Sascal Engneerng Dvson (SED) Robusness Expermens wh Two Varance Componens by Ana Ivelsse Avlés avles@ns.gov Conference

More information

Learning Objectives. Self Organization Map. Hamming Distance(1/5) Introduction. Hamming Distance(3/5) Hamming Distance(2/5) 15/04/2015

Learning Objectives. Self Organization Map. Hamming Distance(1/5) Introduction. Hamming Distance(3/5) Hamming Distance(2/5) 15/04/2015 /4/ Learnng Objecves Self Organzaon Map Learnng whou Exaples. Inroducon. MAXNET 3. Cluserng 4. Feaure Map. Self-organzng Feaure Map 6. Concluson 38 Inroducon. Learnng whou exaples. Daa are npu o he syse

More information

Normal Random Variable and its discriminant functions

Normal Random Variable and its discriminant functions Noral Rando Varable and s dscrnan funcons Oulne Noral Rando Varable Properes Dscrnan funcons Why Noral Rando Varables? Analycally racable Works well when observaon coes for a corruped snle prooype 3 The

More information

New M-Estimator Objective Function. in Simultaneous Equations Model. (A Comparative Study)

New M-Estimator Objective Function. in Simultaneous Equations Model. (A Comparative Study) Inernaonal Mahemacal Forum, Vol. 8, 3, no., 7 - HIKARI Ld, www.m-hkar.com hp://dx.do.org/.988/mf.3.3488 New M-Esmaor Objecve Funcon n Smulaneous Equaons Model (A Comparave Sudy) Ahmed H. Youssef Professor

More information

John Geweke a and Gianni Amisano b a Departments of Economics and Statistics, University of Iowa, USA b European Central Bank, Frankfurt, Germany

John Geweke a and Gianni Amisano b a Departments of Economics and Statistics, University of Iowa, USA b European Central Bank, Frankfurt, Germany Herarchcal Markov Normal Mxure models wh Applcaons o Fnancal Asse Reurns Appendx: Proofs of Theorems and Condonal Poseror Dsrbuons John Geweke a and Gann Amsano b a Deparmens of Economcs and Sascs, Unversy

More information

Variants of Pegasos. December 11, 2009

Variants of Pegasos. December 11, 2009 Inroducon Varans of Pegasos SooWoong Ryu bshboy@sanford.edu December, 009 Youngsoo Cho yc344@sanford.edu Developng a new SVM algorhm s ongong research opc. Among many exng SVM algorhms, we wll focus on

More information

( t) Outline of program: BGC1: Survival and event history analysis Oslo, March-May Recapitulation. The additive regression model

( t) Outline of program: BGC1: Survival and event history analysis Oslo, March-May Recapitulation. The additive regression model BGC1: Survval and even hsory analyss Oslo, March-May 212 Monday May 7h and Tuesday May 8h The addve regresson model Ørnulf Borgan Deparmen of Mahemacs Unversy of Oslo Oulne of program: Recapulaon Counng

More information

Linear Response Theory: The connection between QFT and experiments

Linear Response Theory: The connection between QFT and experiments Phys540.nb 39 3 Lnear Response Theory: The connecon beween QFT and expermens 3.1. Basc conceps and deas Q: ow do we measure he conducvy of a meal? A: we frs nroduce a weak elecrc feld E, and hen measure

More information

( ) () we define the interaction representation by the unitary transformation () = ()

( ) () we define the interaction representation by the unitary transformation () = () Hgher Order Perurbaon Theory Mchael Fowler 3/7/6 The neracon Represenaon Recall ha n he frs par of hs course sequence, we dscussed he chrödnger and Hesenberg represenaons of quanum mechancs here n he chrödnger

More information

TSS = SST + SSE An orthogonal partition of the total SS

TSS = SST + SSE An orthogonal partition of the total SS ANOVA: Topc 4. Orhogonal conrass [ST&D p. 183] H 0 : µ 1 = µ =... = µ H 1 : The mean of a leas one reamen group s dfferen To es hs hypohess, a basc ANOVA allocaes he varaon among reamen means (SST) equally

More information

GMM parameter estimation. Xiaoye Lu CMPS290c Final Project

GMM parameter estimation. Xiaoye Lu CMPS290c Final Project GMM paraeer esaon Xaoye Lu M290c Fnal rojec GMM nroducon Gaussan ure Model obnaon of several gaussan coponens Noaon: For each Gaussan dsrbuon:, s he ean and covarance ar. A GMM h ures(coponens): p ( 2π

More information

UNIVERSITAT AUTÒNOMA DE BARCELONA MARCH 2017 EXAMINATION

UNIVERSITAT AUTÒNOMA DE BARCELONA MARCH 2017 EXAMINATION INTERNATIONAL TRADE T. J. KEHOE UNIVERSITAT AUTÒNOMA DE BARCELONA MARCH 27 EXAMINATION Please answer wo of he hree quesons. You can consul class noes, workng papers, and arcles whle you are workng on he

More information

FI 3103 Quantum Physics

FI 3103 Quantum Physics /9/4 FI 33 Quanum Physcs Aleander A. Iskandar Physcs of Magnesm and Phooncs Research Grou Insu Teknolog Bandung Basc Conces n Quanum Physcs Probably and Eecaon Value Hesenberg Uncerany Prncle Wave Funcon

More information

Chapter 6 DETECTION AND ESTIMATION: Model of digital communication system. Fundamental issues in digital communications are

Chapter 6 DETECTION AND ESTIMATION: Model of digital communication system. Fundamental issues in digital communications are Chaper 6 DCIO AD IMAIO: Fndaenal sses n dgal concaons are. Deecon and. saon Deecon heory: I deals wh he desgn and evalaon of decson ang processor ha observes he receved sgnal and gesses whch parclar sybol

More information

Time-interval analysis of β decay. V. Horvat and J. C. Hardy

Time-interval analysis of β decay. V. Horvat and J. C. Hardy Tme-nerval analyss of β decay V. Horva and J. C. Hardy Work on he even analyss of β decay [1] connued and resuled n he developmen of a novel mehod of bea-decay me-nerval analyss ha produces hghly accurae

More information

Robust and Accurate Cancer Classification with Gene Expression Profiling

Robust and Accurate Cancer Classification with Gene Expression Profiling Robus and Accurae Cancer Classfcaon wh Gene Expresson Proflng (Compuaonal ysems Bology, 2005) Auhor: Hafeng L, Keshu Zhang, ao Jang Oulne Background LDA (lnear dscrmnan analyss) and small sample sze problem

More information

Econ107 Applied Econometrics Topic 5: Specification: Choosing Independent Variables (Studenmund, Chapter 6)

Econ107 Applied Econometrics Topic 5: Specification: Choosing Independent Variables (Studenmund, Chapter 6) Econ7 Appled Economercs Topc 5: Specfcaon: Choosng Independen Varables (Sudenmund, Chaper 6 Specfcaon errors ha we wll deal wh: wrong ndependen varable; wrong funconal form. Ths lecure deals wh wrong ndependen

More information

THE PUBLISHING HOUSE PROCEEDINGS OF THE ROMANIAN ACADEMY, Series A, OF THE ROMANIAN ACADEMY Volume 9, Number 1/2008, pp

THE PUBLISHING HOUSE PROCEEDINGS OF THE ROMANIAN ACADEMY, Series A, OF THE ROMANIAN ACADEMY Volume 9, Number 1/2008, pp THE PUBLISHING HOUSE PROCEEDINGS OF THE ROMNIN CDEMY, Seres, OF THE ROMNIN CDEMY Volue 9, Nuber /008, pp. 000 000 ON CIMMINO'S REFLECTION LGORITHM Consann POP Ovdus Unversy of Consana, Roana, E-al: cpopa@unv-ovdus.ro

More information

Solution in semi infinite diffusion couples (error function analysis)

Solution in semi infinite diffusion couples (error function analysis) Soluon n sem nfne dffuson couples (error funcon analyss) Le us consder now he sem nfne dffuson couple of wo blocks wh concenraon of and I means ha, n a A- bnary sysem, s bondng beween wo blocks made of

More information

In the complete model, these slopes are ANALYSIS OF VARIANCE FOR THE COMPLETE TWO-WAY MODEL. (! i+1 -! i ) + [(!") i+1,q - [(!

In the complete model, these slopes are ANALYSIS OF VARIANCE FOR THE COMPLETE TWO-WAY MODEL. (! i+1 -! i ) + [(!) i+1,q - [(! ANALYSIS OF VARIANCE FOR THE COMPLETE TWO-WAY MODEL The frs hng o es n wo-way ANOVA: Is here neracon? "No neracon" means: The man effecs model would f. Ths n urn means: In he neracon plo (wh A on he horzonal

More information

Bayesian Model Selection for Structural Break Models *

Bayesian Model Selection for Structural Break Models * Baesan Model Selecon for Srucural Brea Models * Andrew T. Levn Federal eserve Board Jere M. Pger Unvers of Oregon Frs Verson: Noveber 005 Ths verson: Aprl 007 Absrac: We ae a Baesan approach o odel selecon

More information

CS434a/541a: Pattern Recognition Prof. Olga Veksler. Lecture 4

CS434a/541a: Pattern Recognition Prof. Olga Veksler. Lecture 4 CS434a/54a: Paern Recognon Prof. Olga Veksler Lecure 4 Oulne Normal Random Varable Properes Dscrmnan funcons Why Normal Random Varables? Analycally racable Works well when observaon comes form a corruped

More information

Math 128b Project. Jude Yuen

Math 128b Project. Jude Yuen Mah 8b Proec Jude Yuen . Inroducon Le { Z } be a sequence of observed ndependen vecor varables. If he elemens of Z have a on normal dsrbuon hen { Z } has a mean vecor Z and a varancecovarance marx z. Geomercally

More information

Appendix H: Rarefaction and extrapolation of Hill numbers for incidence data

Appendix H: Rarefaction and extrapolation of Hill numbers for incidence data Anne Chao Ncholas J Goell C seh lzabeh L ander K Ma Rober K Colwell and Aaron M llson 03 Rarefacon and erapolaon wh ll numbers: a framewor for samplng and esmaon n speces dversy sudes cology Monographs

More information

CS286.2 Lecture 14: Quantum de Finetti Theorems II

CS286.2 Lecture 14: Quantum de Finetti Theorems II CS286.2 Lecure 14: Quanum de Fne Theorems II Scrbe: Mara Okounkova 1 Saemen of he heorem Recall he las saemen of he quanum de Fne heorem from he prevous lecure. Theorem 1 Quanum de Fne). Le ρ Dens C 2

More information

Lecture 6: Learning for Control (Generalised Linear Regression)

Lecture 6: Learning for Control (Generalised Linear Regression) Lecure 6: Learnng for Conrol (Generalsed Lnear Regresson) Conens: Lnear Mehods for Regresson Leas Squares, Gauss Markov heorem Recursve Leas Squares Lecure 6: RLSC - Prof. Sehu Vjayakumar Lnear Regresson

More information

Volatility Interpolation

Volatility Interpolation Volaly Inerpolaon Prelmnary Verson March 00 Jesper Andreasen and Bran Huge Danse Mares, Copenhagen wan.daddy@danseban.com brno@danseban.com Elecronc copy avalable a: hp://ssrn.com/absrac=69497 Inro Local

More information

. The geometric multiplicity is dim[ker( λi. number of linearly independent eigenvectors associated with this eigenvalue.

. The geometric multiplicity is dim[ker( λi. number of linearly independent eigenvectors associated with this eigenvalue. Lnear Algebra Lecure # Noes We connue wh he dscusson of egenvalues, egenvecors, and dagonalzably of marces We wan o know, n parcular wha condons wll assure ha a marx can be dagonalzed and wha he obsrucons

More information

Lecture VI Regression

Lecture VI Regression Lecure VI Regresson (Lnear Mehods for Regresson) Conens: Lnear Mehods for Regresson Leas Squares, Gauss Markov heorem Recursve Leas Squares Lecure VI: MLSC - Dr. Sehu Vjayakumar Lnear Regresson Model M

More information

Lecture 18: The Laplace Transform (See Sections and 14.7 in Boas)

Lecture 18: The Laplace Transform (See Sections and 14.7 in Boas) Lecure 8: The Lalace Transform (See Secons 88- and 47 n Boas) Recall ha our bg-cure goal s he analyss of he dfferenal equaon, ax bx cx F, where we emloy varous exansons for he drvng funcon F deendng on

More information

GENERATING CERTAIN QUINTIC IRREDUCIBLE POLYNOMIALS OVER FINITE FIELDS. Youngwoo Ahn and Kitae Kim

GENERATING CERTAIN QUINTIC IRREDUCIBLE POLYNOMIALS OVER FINITE FIELDS. Youngwoo Ahn and Kitae Kim Korean J. Mah. 19 (2011), No. 3, pp. 263 272 GENERATING CERTAIN QUINTIC IRREDUCIBLE POLYNOMIALS OVER FINITE FIELDS Youngwoo Ahn and Kae Km Absrac. In he paper [1], an explc correspondence beween ceran

More information

Let s treat the problem of the response of a system to an applied external force. Again,

Let s treat the problem of the response of a system to an applied external force. Again, Page 33 QUANTUM LNEAR RESPONSE FUNCTON Le s rea he problem of he response of a sysem o an appled exernal force. Agan, H() H f () A H + V () Exernal agen acng on nernal varable Hamlonan for equlbrum sysem

More information

Two-Step versus Simultaneous Estimation of Survey-Non-Sampling Error and True Value Components of Small Area Sample Estimators

Two-Step versus Simultaneous Estimation of Survey-Non-Sampling Error and True Value Components of Small Area Sample Estimators Two-Sep versus Sulaneous Esaon of Survey-Non-Saplng Error and True Value Coponens of Sall rea Saple Esaors a PVB Sway, TS Zeran b c and JS Meha a,b Bureau of Labor Sascs, Roo 4985, Massachuses venue, NE,

More information

Comparison of Differences between Power Means 1

Comparison of Differences between Power Means 1 In. Journal of Mah. Analyss, Vol. 7, 203, no., 5-55 Comparson of Dfferences beween Power Means Chang-An Tan, Guanghua Sh and Fe Zuo College of Mahemacs and Informaon Scence Henan Normal Unversy, 453007,

More information

Existence and Uniqueness Results for Random Impulsive Integro-Differential Equation

Existence and Uniqueness Results for Random Impulsive Integro-Differential Equation Global Journal of Pure and Appled Mahemacs. ISSN 973-768 Volume 4, Number 6 (8), pp. 89-87 Research Inda Publcaons hp://www.rpublcaon.com Exsence and Unqueness Resuls for Random Impulsve Inegro-Dfferenal

More information

Modeling of Combined Deterioration of Concrete Structures by Competing Hazard Model

Modeling of Combined Deterioration of Concrete Structures by Competing Hazard Model Modelng of Cobned Deeroraon of Concree Srucures by Copeng Hazard Model Kyoyuk KAITO Assocae Professor Froner Research Cener Osaka Unv., Osaka, apan kao@ga.eng.osaka-u.ac.p Kyoyuk KAITO, born 97, receved

More information

On One Analytic Method of. Constructing Program Controls

On One Analytic Method of. Constructing Program Controls Appled Mahemacal Scences, Vol. 9, 05, no. 8, 409-407 HIKARI Ld, www.m-hkar.com hp://dx.do.org/0.988/ams.05.54349 On One Analyc Mehod of Consrucng Program Conrols A. N. Kvko, S. V. Chsyakov and Yu. E. Balyna

More information

THE PREDICTION OF COMPETITIVE ENVIRONMENT IN BUSINESS

THE PREDICTION OF COMPETITIVE ENVIRONMENT IN BUSINESS THE PREICTION OF COMPETITIVE ENVIRONMENT IN BUSINESS INTROUCTION The wo dmensonal paral dfferenal equaons of second order can be used for he smulaon of compeve envronmen n busness The arcle presens he

More information

Response of MDOF systems

Response of MDOF systems Response of MDOF syses Degree of freedo DOF: he nu nuber of ndependen coordnaes requred o deerne copleely he posons of all pars of a syse a any nsan of e. wo DOF syses hree DOF syses he noral ode analyss

More information

Long Term Power Load Combination Forecasting Based on Chaos-Fractal Theory in Beijing

Long Term Power Load Combination Forecasting Based on Chaos-Fractal Theory in Beijing JAGUO ZHOU e al: LOG TERM POWER LOAD COMBIATIO FORECASTIG BASED O CHAOS Long Ter Power Load Cobnaon Forecasng Based on Chaos-Fracal Theory n Bejng Janguo Zhou,We Lu,*,Qang Song School of Econocs and Manageen

More information

. The geometric multiplicity is dim[ker( λi. A )], i.e. the number of linearly independent eigenvectors associated with this eigenvalue.

. The geometric multiplicity is dim[ker( λi. A )], i.e. the number of linearly independent eigenvectors associated with this eigenvalue. Mah E-b Lecure #0 Noes We connue wh he dscusson of egenvalues, egenvecors, and dagonalzably of marces We wan o know, n parcular wha condons wll assure ha a marx can be dagonalzed and wha he obsrucons are

More information

Density Matrix Description of NMR BCMB/CHEM 8190

Density Matrix Description of NMR BCMB/CHEM 8190 Densy Marx Descrpon of NMR BCMBCHEM 89 Operaors n Marx Noaon Alernae approach o second order specra: ask abou x magnezaon nsead of energes and ranson probables. If we say wh one bass se, properes vary

More information

Capacity of TWSC Intersection with Multilane Approaches

Capacity of TWSC Intersection with Multilane Approaches Avalable onlne a www.scencedrec.co roceda Socal and Behavoral Scences 16 (2011) 664 675 6 h Inernaonal Syposu on Hghway apacy and Qualy of Servce Sochol Sweden June 28 July 1 2011 apacy of TWS Inersecon

More information

Survival Analysis and Reliability. A Note on the Mean Residual Life Function of a Parallel System

Survival Analysis and Reliability. A Note on the Mean Residual Life Function of a Parallel System Communcaons n Sascs Theory and Mehods, 34: 475 484, 2005 Copyrgh Taylor & Francs, Inc. ISSN: 0361-0926 prn/1532-415x onlne DOI: 10.1081/STA-200047430 Survval Analyss and Relably A Noe on he Mean Resdual

More information

Fall 2010 Graduate Course on Dynamic Learning

Fall 2010 Graduate Course on Dynamic Learning Fall 200 Graduae Course on Dynamc Learnng Chaper 4: Parcle Flers Sepember 27, 200 Byoung-Tak Zhang School of Compuer Scence and Engneerng & Cognve Scence and Bran Scence Programs Seoul aonal Unversy hp://b.snu.ac.kr/~bzhang/

More information

RELATIONSHIP BETWEEN VOLATILITY AND TRADING VOLUME: THE CASE OF HSI STOCK RETURNS DATA

RELATIONSHIP BETWEEN VOLATILITY AND TRADING VOLUME: THE CASE OF HSI STOCK RETURNS DATA RELATIONSHIP BETWEEN VOLATILITY AND TRADING VOLUME: THE CASE OF HSI STOCK RETURNS DATA Mchaela Chocholaá Unversy of Economcs Braslava, Slovaka Inroducon (1) one of he characersc feaures of sock reurns

More information

Graduate Macroeconomics 2 Problem set 5. - Solutions

Graduate Macroeconomics 2 Problem set 5. - Solutions Graduae Macroeconomcs 2 Problem se. - Soluons Queson 1 To answer hs queson we need he frms frs order condons and he equaon ha deermnes he number of frms n equlbrum. The frms frs order condons are: F K

More information

A Modified Genetic Algorithm Comparable to Quantum GA

A Modified Genetic Algorithm Comparable to Quantum GA A Modfed Genec Algorh Coparable o Quanu GA Tahereh Kahookar Toos Ferdows Unversy of Mashhad _k_oos@wal.u.ac.r Habb Rajab Mashhad Ferdows Unversy of Mashhad h_rajab@ferdows.u.ac.r Absrac: Recenly, researchers

More information

Online Supplement for Dynamic Multi-Technology. Production-Inventory Problem with Emissions Trading

Online Supplement for Dynamic Multi-Technology. Production-Inventory Problem with Emissions Trading Onlne Supplemen for Dynamc Mul-Technology Producon-Invenory Problem wh Emssons Tradng by We Zhang Zhongsheng Hua Yu Xa and Baofeng Huo Proof of Lemma For any ( qr ) Θ s easy o verfy ha he lnear programmng

More information

Optimal environmental charges under imperfect compliance

Optimal environmental charges under imperfect compliance ISSN 1 746-7233, England, UK World Journal of Modellng and Smulaon Vol. 4 (28) No. 2, pp. 131-139 Opmal envronmenal charges under mperfec complance Dajn Lu 1, Ya Wang 2 Tazhou Insue of Scence and Technology,

More information

Cubic Bezier Homotopy Function for Solving Exponential Equations

Cubic Bezier Homotopy Function for Solving Exponential Equations Penerb Journal of Advanced Research n Compung and Applcaons ISSN (onlne: 46-97 Vol. 4, No.. Pages -8, 6 omoopy Funcon for Solvng Eponenal Equaons S. S. Raml *,,. Mohamad Nor,a, N. S. Saharzan,b and M.

More information

Clustering (Bishop ch 9)

Clustering (Bishop ch 9) Cluserng (Bshop ch 9) Reference: Daa Mnng by Margare Dunham (a slde source) 1 Cluserng Cluserng s unsupervsed learnng, here are no class labels Wan o fnd groups of smlar nsances Ofen use a dsance measure

More information

( ) [ ] MAP Decision Rule

( ) [ ] MAP Decision Rule Announcemens Bayes Decson Theory wh Normal Dsrbuons HW0 due oday HW o be assgned soon Proec descrpon posed Bomercs CSE 90 Lecure 4 CSE90, Sprng 04 CSE90, Sprng 04 Key Probables 4 ω class label X feaure

More information

A TWO-LEVEL LOAN PORTFOLIO OPTIMIZATION PROBLEM

A TWO-LEVEL LOAN PORTFOLIO OPTIMIZATION PROBLEM Proceedngs of he 2010 Wner Sulaon Conference B. Johansson, S. Jan, J. Monoya-Torres, J. Hugan, and E. Yücesan, eds. A TWO-LEVEL LOAN PORTFOLIO OPTIMIZATION PROBLEM JanQang Hu Jun Tong School of Manageen

More information

Notes on the stability of dynamic systems and the use of Eigen Values.

Notes on the stability of dynamic systems and the use of Eigen Values. Noes on he sabl of dnamc ssems and he use of Egen Values. Source: Macro II course noes, Dr. Davd Bessler s Tme Seres course noes, zarads (999) Ineremporal Macroeconomcs chaper 4 & Techncal ppend, and Hamlon

More information

F-Tests and Analysis of Variance (ANOVA) in the Simple Linear Regression Model. 1. Introduction

F-Tests and Analysis of Variance (ANOVA) in the Simple Linear Regression Model. 1. Introduction ECOOMICS 35* -- OTE 9 ECO 35* -- OTE 9 F-Tess and Analyss of Varance (AOVA n he Smple Lnear Regresson Model Inroducon The smple lnear regresson model s gven by he followng populaon regresson equaon, or

More information

[Link to MIT-Lab 6P.1 goes here.] After completing the lab, fill in the following blanks: Numerical. Simulation s Calculations

[Link to MIT-Lab 6P.1 goes here.] After completing the lab, fill in the following blanks: Numerical. Simulation s Calculations Chaper 6: Ordnary Leas Squares Esmaon Procedure he Properes Chaper 6 Oulne Cln s Assgnmen: Assess he Effec of Sudyng on Quz Scores Revew o Regresson Model o Ordnary Leas Squares () Esmaon Procedure o he

More information

Chapter Lagrangian Interpolation

Chapter Lagrangian Interpolation Chaper 5.4 agrangan Inerpolaon Afer readng hs chaper you should be able o:. dere agrangan mehod of nerpolaon. sole problems usng agrangan mehod of nerpolaon and. use agrangan nerpolans o fnd deraes and

More information

5th International Conference on Advanced Design and Manufacturing Engineering (ICADME 2015)

5th International Conference on Advanced Design and Manufacturing Engineering (ICADME 2015) 5h Inernaonal onference on Advanced Desgn and Manufacurng Engneerng (IADME 5 The Falure Rae Expermenal Sudy of Specal N Machne Tool hunshan He, a, *, La Pan,b and Bng Hu 3,c,,3 ollege of Mechancal and

More information

Robustness of DEWMA versus EWMA Control Charts to Non-Normal Processes

Robustness of DEWMA versus EWMA Control Charts to Non-Normal Processes Journal of Modern Appled Sascal Mehods Volume Issue Arcle 8 5--3 Robusness of D versus Conrol Chars o Non- Processes Saad Saeed Alkahan Performance Measuremen Cener of Governmen Agences, Insue of Publc

More information

DEEP UNFOLDING FOR MULTICHANNEL SOURCE SEPARATION SUPPLEMENTARY MATERIAL

DEEP UNFOLDING FOR MULTICHANNEL SOURCE SEPARATION SUPPLEMENTARY MATERIAL DEEP UNFOLDING FOR MULTICHANNEL SOURCE SEPARATION SUPPLEMENTARY MATERIAL Sco Wsdom, John Hershey 2, Jonahan Le Roux 2, and Shnj Waanabe 2 Deparmen o Elecrcal Engneerng, Unversy o Washngon, Seale, WA, USA

More information

Mechanics Physics 151

Mechanics Physics 151 Mechancs Physcs 5 Lecure 9 Hamlonan Equaons of Moon (Chaper 8) Wha We Dd Las Tme Consruced Hamlonan formalsm H ( q, p, ) = q p L( q, q, ) H p = q H q = p H = L Equvalen o Lagrangan formalsm Smpler, bu

More information

Standard Error of Technical Cost Incorporating Parameter Uncertainty

Standard Error of Technical Cost Incorporating Parameter Uncertainty Sandard rror of echncal Cos Incorporang Parameer Uncerany Chrsopher Moron Insurance Ausrala Group Presened o he Acuares Insue General Insurance Semnar 3 ovember 0 Sydney hs paper has been prepared for

More information

A HIERARCHICAL KALMAN FILTER

A HIERARCHICAL KALMAN FILTER A HIERARCHICAL KALMAN FILER Greg aylor aylor Fry Consulng Acuares Level 8, 3 Clarence Sree Sydney NSW Ausrala Professoral Assocae, Cenre for Acuaral Sudes Faculy of Economcs and Commerce Unversy of Melbourne

More information

Outline. Probabilistic Model Learning. Probabilistic Model Learning. Probabilistic Model for Time-series Data: Hidden Markov Model

Outline. Probabilistic Model Learning. Probabilistic Model Learning. Probabilistic Model for Time-series Data: Hidden Markov Model Probablsc Model for Tme-seres Daa: Hdden Markov Model Hrosh Mamsuka Bonformacs Cener Kyoo Unversy Oulne Three Problems for probablsc models n machne learnng. Compung lkelhood 2. Learnng 3. Parsng (predcon

More information

Mechanics Physics 151

Mechanics Physics 151 Mechancs Physcs 5 Lecure 9 Hamlonan Equaons of Moon (Chaper 8) Wha We Dd Las Tme Consruced Hamlonan formalsm Hqp (,,) = qp Lqq (,,) H p = q H q = p H L = Equvalen o Lagrangan formalsm Smpler, bu wce as

More information

CHAPTER 10: LINEAR DISCRIMINATION

CHAPTER 10: LINEAR DISCRIMINATION CHAPER : LINEAR DISCRIMINAION Dscrmnan-based Classfcaon 3 In classfcaon h K classes (C,C,, C k ) We defned dscrmnan funcon g j (), j=,,,k hen gven an es eample, e chose (predced) s class label as C f g

More information

Ordinary Differential Equations in Neuroscience with Matlab examples. Aim 1- Gain understanding of how to set up and solve ODE s

Ordinary Differential Equations in Neuroscience with Matlab examples. Aim 1- Gain understanding of how to set up and solve ODE s Ordnary Dfferenal Equaons n Neuroscence wh Malab eamples. Am - Gan undersandng of how o se up and solve ODE s Am Undersand how o se up an solve a smple eample of he Hebb rule n D Our goal a end of class

More information

January Examinations 2012

January Examinations 2012 Page of 5 EC79 January Examnaons No. of Pages: 5 No. of Quesons: 8 Subjec ECONOMICS (POSTGRADUATE) Tle of Paper EC79 QUANTITATIVE METHODS FOR BUSINESS AND FINANCE Tme Allowed Two Hours ( hours) Insrucons

More information

2. SPATIALLY LAGGED DEPENDENT VARIABLES

2. SPATIALLY LAGGED DEPENDENT VARIABLES 2. SPATIALLY LAGGED DEPENDENT VARIABLES In hs chaper, we descrbe a sascal model ha ncorporaes spaal dependence explcly by addng a spaally lagged dependen varable y on he rgh-hand sde of he regresson equaon.

More information

Comparison of Supervised & Unsupervised Learning in βs Estimation between Stocks and the S&P500

Comparison of Supervised & Unsupervised Learning in βs Estimation between Stocks and the S&P500 Comparson of Supervsed & Unsupervsed Learnng n βs Esmaon beween Socks and he S&P500 J. We, Y. Hassd, J. Edery, A. Becker, Sanford Unversy T I. INTRODUCTION HE goal of our proec s o analyze he relaonshps

More information

Fall 2009 Social Sciences 7418 University of Wisconsin-Madison. Problem Set 2 Answers (4) (6) di = D (10)

Fall 2009 Social Sciences 7418 University of Wisconsin-Madison. Problem Set 2 Answers (4) (6) di = D (10) Publc Affars 974 Menze D. Chnn Fall 2009 Socal Scences 7418 Unversy of Wsconsn-Madson Problem Se 2 Answers Due n lecure on Thursday, November 12. " Box n" your answers o he algebrac quesons. 1. Consder

More information

Part II CONTINUOUS TIME STOCHASTIC PROCESSES

Part II CONTINUOUS TIME STOCHASTIC PROCESSES Par II CONTINUOUS TIME STOCHASTIC PROCESSES 4 Chaper 4 For an advanced analyss of he properes of he Wener process, see: Revus D and Yor M: Connuous marngales and Brownan Moon Karazas I and Shreve S E:

More information

Fitting a transformation: Feature based alignment May 1 st, 2018

Fitting a transformation: Feature based alignment May 1 st, 2018 5//8 Fng a ransforaon: Feaure based algnen Ma s, 8 Yong Jae Lee UC Davs Las e: Deforable conours a.k.a. acve conours, snakes Gven: nal conour (odel) near desred objec Goal: evolve he conour o f eac objec

More information

CHAPTER 5: MULTIVARIATE METHODS

CHAPTER 5: MULTIVARIATE METHODS CHAPER 5: MULIVARIAE MEHODS Mulvarae Daa 3 Mulple measuremens (sensors) npus/feaures/arbues: -varae N nsances/observaons/eamples Each row s an eample Each column represens a feaure X a b correspons o he

More information

Performance Analysis for a Network having Standby Redundant Unit with Waiting in Repair

Performance Analysis for a Network having Standby Redundant Unit with Waiting in Repair TECHNI Inernaonal Journal of Compung Scence Communcaon Technologes VOL.5 NO. July 22 (ISSN 974-3375 erformance nalyss for a Nework havng Sby edundan Un wh ang n epar Jendra Sngh 2 abns orwal 2 Deparmen

More information

Comb Filters. Comb Filters

Comb Filters. Comb Filters The smple flers dscussed so far are characered eher by a sngle passband and/or a sngle sopband There are applcaons where flers wh mulple passbands and sopbands are requred Thecomb fler s an example of

More information

Tight results for Next Fit and Worst Fit with resource augmentation

Tight results for Next Fit and Worst Fit with resource augmentation Tgh resuls for Nex F and Wors F wh resource augmenaon Joan Boyar Leah Epsen Asaf Levn Asrac I s well known ha he wo smple algorhms for he classc n packng prolem, NF and WF oh have an approxmaon rao of

More information

Statistical matching using fractional imputation

Statistical matching using fractional imputation Sascs Publcaons Sascs 6-06 Sascal achng usng fraconal puaon Jae Kwang K Iowa Sae Unvers, jk@asae.edu El J. Berg Iowa Sae Unvers, elb@asae.edu Taesung Park Seoul Naonal Unvers Follow hs and addonal works

More information

グラフィカルモデルによる推論 確率伝搬法 (2) Kenji Fukumizu The Institute of Statistical Mathematics 計算推論科学概論 II (2010 年度, 後期 )

グラフィカルモデルによる推論 確率伝搬法 (2) Kenji Fukumizu The Institute of Statistical Mathematics 計算推論科学概論 II (2010 年度, 後期 ) グラフィカルモデルによる推論 確率伝搬法 Kenj Fukuzu he Insue of Sascal Maheacs 計算推論科学概論 II 年度 後期 Inference on Hdden Markov Model Inference on Hdden Markov Model Revew: HMM odel : hdden sae fne Inference Coue... for any Naïve

More information

Tools for Analysis of Accelerated Life and Degradation Test Data

Tools for Analysis of Accelerated Life and Degradation Test Data Acceleraed Sress Tesng and Relably Tools for Analyss of Acceleraed Lfe and Degradaon Tes Daa Presened by: Reuel Smh Unversy of Maryland College Park smhrc@umd.edu Sepember-5-6 Sepember 28-30 206, Pensacola

More information

Advanced Macroeconomics II: Exchange economy

Advanced Macroeconomics II: Exchange economy Advanced Macroeconomcs II: Exchange economy Krzyszof Makarsk 1 Smple deermnsc dynamc model. 1.1 Inroducon Inroducon Smple deermnsc dynamc model. Defnons of equlbrum: Arrow-Debreu Sequenal Recursve Equvalence

More information

A DECOMPOSITION METHOD FOR SOLVING DIFFUSION EQUATIONS VIA LOCAL FRACTIONAL TIME DERIVATIVE

A DECOMPOSITION METHOD FOR SOLVING DIFFUSION EQUATIONS VIA LOCAL FRACTIONAL TIME DERIVATIVE S13 A DECOMPOSITION METHOD FOR SOLVING DIFFUSION EQUATIONS VIA LOCAL FRACTIONAL TIME DERIVATIVE by Hossen JAFARI a,b, Haleh TAJADODI c, and Sarah Jane JOHNSTON a a Deparen of Maheacal Scences, Unversy

More information

1 Constant Real Rate C 1

1 Constant Real Rate C 1 Consan Real Rae. Real Rae of Inees Suppose you ae equally happy wh uns of he consumpon good oday o 5 uns of he consumpon good n peod s me. C 5 Tha means you ll be pepaed o gve up uns oday n eun fo 5 uns

More information

AT&T Labs Research, Shannon Laboratory, 180 Park Avenue, Room A279, Florham Park, NJ , USA

AT&T Labs Research, Shannon Laboratory, 180 Park Avenue, Room A279, Florham Park, NJ , USA Machne Learnng, 43, 65 91, 001 c 001 Kluwer Acadec Publshers. Manufacured n The Neherlands. Drfng Gaes ROBERT E. SCHAPIRE schapre@research.a.co AT&T Labs Research, Shannon Laboraory, 180 Park Avenue, Roo

More information

Midterm Exam. Thursday, April hour, 15 minutes

Midterm Exam. Thursday, April hour, 15 minutes Economcs of Grow, ECO560 San Francsco Sae Unvers Mcael Bar Sprng 04 Mderm Exam Tursda, prl 0 our, 5 mnues ame: Insrucons. Ts s closed boo, closed noes exam.. o calculaors of an nd are allowed. 3. Sow all

More information

Changeovers. Department of Chemical Engineering, Carnegie Mellon University, Pittsburgh, PA 15213, USA

Changeovers. Department of Chemical Engineering, Carnegie Mellon University, Pittsburgh, PA 15213, USA wo ew Connuous-e odels for he Schedulng of ulsage Bach Plans wh Sequence Dependen Changeovers Pedro. Casro * gnaco E. Grossann and Auguso Q. ovas Deparaeno de odelação e Sulação de Processos E 649-038

More information

Density Matrix Description of NMR BCMB/CHEM 8190

Density Matrix Description of NMR BCMB/CHEM 8190 Densy Marx Descrpon of NMR BCMBCHEM 89 Operaors n Marx Noaon If we say wh one bass se, properes vary only because of changes n he coeffcens weghng each bass se funcon x = h< Ix > - hs s how we calculae

More information

FTCS Solution to the Heat Equation

FTCS Solution to the Heat Equation FTCS Soluon o he Hea Equaon ME 448/548 Noes Gerald Reckenwald Porland Sae Unversy Deparmen of Mechancal Engneerng gerry@pdxedu ME 448/548: FTCS Soluon o he Hea Equaon Overvew Use he forward fne d erence

More information

Introduction to Numerical Analysis. In this lesson you will be taken through a pair of techniques that will be used to solve the equations of.

Introduction to Numerical Analysis. In this lesson you will be taken through a pair of techniques that will be used to solve the equations of. Inroducion o Nuerical Analysis oion In his lesson you will be aen hrough a pair of echniques ha will be used o solve he equaions of and v dx d a F d for siuaions in which F is well nown, and he iniial

More information

II. Light is a Ray (Geometrical Optics)

II. Light is a Ray (Geometrical Optics) II Lgh s a Ray (Geomercal Opcs) IIB Reflecon and Refracon Hero s Prncple of Leas Dsance Law of Reflecon Hero of Aleandra, who lved n he 2 nd cenury BC, posulaed he followng prncple: Prncple of Leas Dsance:

More information

Time Truncated Sampling Plan under Hybrid Exponential Distribution

Time Truncated Sampling Plan under Hybrid Exponential Distribution 18118118118118117 Journal of Unceran Syses Vol.1, No.3, pp.181-197, 16 Onlne a: www.jus.org.uk Te Truncaed Saplng Plan under Hybrd Exponenal Dsrbuon S. Sapah 1,, S. M. Lalha 1 Deparen of Sascs, Unversy

More information

Transcription: Messenger RNA, mrna, is produced and transported to Ribosomes

Transcription: Messenger RNA, mrna, is produced and transported to Ribosomes Quanave Cenral Dogma I Reference hp//book.bonumbers.org Inaon ranscrpon RNA polymerase and ranscrpon Facor (F) s bnds o promoer regon of DNA ranscrpon Meenger RNA, mrna, s produced and ranspored o Rbosomes

More information

Sklar: Sections (4.4.2 is not covered).

Sklar: Sections (4.4.2 is not covered). COSC 44: Dgal Councaons Insrucor: Dr. Ar Asf Deparen of Copuer Scence and Engneerng York Unversy Handou # 6: Bandpass Modulaon opcs:. Phasor Represenaon. Dgal Modulaon Schees: PSK FSK ASK APK ASK/FSK)

More information

(,,, ) (,,, ). In addition, there are three other consumers, -2, -1, and 0. Consumer -2 has the utility function

(,,, ) (,,, ). In addition, there are three other consumers, -2, -1, and 0. Consumer -2 has the utility function MACROECONOMIC THEORY T J KEHOE ECON 87 SPRING 5 PROBLEM SET # Conder an overlappng generaon economy le ha n queon 5 on problem e n whch conumer lve for perod The uly funcon of he conumer born n perod,

More information

Introduction to Boosting

Introduction to Boosting Inroducon o Boosng Cynha Rudn PACM, Prnceon Unversy Advsors Ingrd Daubeches and Rober Schapre Say you have a daabase of news arcles, +, +, -, -, +, +, -, -, +, +, -, -, +, +, -, + where arcles are labeled

More information

e-journal Reliability: Theory& Applications No 2 (Vol.2) Vyacheslav Abramov

e-journal Reliability: Theory& Applications No 2 (Vol.2) Vyacheslav Abramov June 7 e-ournal Relably: Theory& Applcaons No (Vol. CONFIDENCE INTERVALS ASSOCIATED WITH PERFORMANCE ANALYSIS OF SYMMETRIC LARGE CLOSED CLIENT/SERVER COMPUTER NETWORKS Absrac Vyacheslav Abramov School

More information