University of Wollongong. Research Online

Size: px
Start display at page:

Download "University of Wollongong. Research Online"

Transcription

1 Unversty of Wollongong Research Onlne Centre for Statstcal & Survey Methodology Workng Paper Seres Faculty of Engneerng and Informaton Scences 2009 Borrowng strength over space n small area estmaton: Comparng parametrc, semparametrc and non-parametrc random effects and M-quantle small area models R. Chambers Unversty of Wollongong, ray@uow.edu.au N. Tzavds Unversty of Manchester N. Salvat Unversty of Psa Recommended Ctaton Chambers, R.; Tzavds, N.; and Salvat, N., Borrowng strength over space n small area estmaton: Comparng parametrc, semparametrc and non-parametrc random effects and M-quantle small area models, Centre for Statstcal and Survey Methodology, Unversty of Wollongong, Workng Paper 12-09, 2009, 13p. Research Onlne s the open access nsttutonal repostory for the Unversty of Wollongong. For further nformaton contact the UOW Lbrary: research-pubs@uow.edu.au

2 Centre for Statstcal and Survey Methodology The Unversty of Wollongong Workng Paper Borrowng strength over space n small area estmaton: Comparng parametrc, sem-parametrc and non-parametrc random effects and M- quantle small area models Ray Chambers, Nkos Tzavds, Ncola Salvat Copyrght 2008 by the Centre for Statstcal & Survey Methodology, UOW. Work n progress, no part of ths paper may be reproduced wthout permsson from the Centre. Centre for Statstcal & Survey Methodology, Unversty of Wollongong, Wollongong NSW Phone , Fax Emal: anca@uow.edu.au

3 Borrowng strength over space n small area estmaton: Comparng parametrc, sem-parametrc and non-parametrc random effects and M-quantle small area models Ray Chambers 1, Nkos Tzavds 2, Ncola Salvat 3 Centre for Statstcal and Survey Methodology School of Mathematcs and Appled Statstcs, Unversty of Wollongong 1 Centre for Census and Survey Research, Unversty of Manchester 2 Dpartmento d Statstca e Matematca Applcata all Economa, Unverstà d Psa, Italy 3 Abstract In recent years there have been sgnfcant developments n model-based small area methods that ncorporate spatal nformaton n an attempt to mprove the effcency of small area estmates by borrowng strength over space. A popular approach parametrcally models spatal correlaton n area effects usng Smultaneous Autoregressve (SAR) random effects models. An alternatve approach ncorporates the spatal nformaton va M-quantle Geographcally Weghted Regresson (GWR), whch fts a local model to the M-quantles of the condtonal dstrbuton of the outcome varable gven the covarates. A further approach uses splne approxmatons to ft nonparametrc unt level nested error regresson and M-quantle regresson models that reflect spatal varaton n the data and then uses these nonparametrc models for small area estmaton. In ths presentaton we contrast the performance of these alternatve small area models usng data wth geographcal nformaton. We also examne how these models perform when estmaton s for out of sample areas.e. areas wth zero sample, and dscuss ssues related to estmaton of mean squared error of the resultng small area estmators. Our analyss s llustrated usng smulatons based on data from the U.S. Envronmental Protecton Agency s Envronmental Montorng and Assessment Program. Keywords: Spatal nformaton; Robust regresson; Iteratvely Reweghted Least Squares; Nonparametrc smoothng. 1 Introducton In recent years there have been sgnfcant developments n model-based small area estmaton. The most popular approach to small area estmaton employs random effects models for estmatng doman specfc parameters (Rao, 2003). An alternatve approach to small area estmaton that relaxes the parametrc assumptons of random effects models by employng M-quantle models was recently proposed by Chambers & Tzavds (2006). Typcally, random effects models assume ndependence of the random area effects. Ths ndependence assumpton s also mplct n M-quantle small area models. In economc, envronmental and epdemologcal applcatons, however, observatons that are spatally close may be more related than observatons that are further apart. Ths spatal correlaton can be accounted for by extendng the random effects model to allow for spatally correlated area effects usng, for example, a Smultaneous Autoregressve (SAR) model (Anseln, 1988; Cresse, 1993). The applcaton of Smultaneous Autoregressve models to small area estmaton enables researchers to borrow strength over space and hence mprove the precson of small area estmates. In ths context, Sngh et al. (2005) and Prates & Salvat (2008) proposed the use of the Spatal Emprcal Best Lnear Unbased Predctor (SEBLUP). SAR models allow for spatal correlaton n the error structure. An alternatve approach to ncorporatng the spatal nformaton n the regresson model s by assumng that the regresson coeffcents vary spatally across the geography of nterest. Geographcally Weghted Regresson (GWR) (Brundson et al., 1996; Fotherngham et al., 1997, 2002; Yu & Wu, 2004) extends the tradtonal regresson model by allowng local rather than global parameters to be estmated. That s, GWR drectly models spatally non-statonarty n the mean structure of the model. In a recent paper Salvat et al. (2008) proposed an M-quantle GWR small area model. In dong so, the authors frst proposed an extenson to the GWR model, M-quantle GWR model,.e. a locally robust model for the M-quantles of the condtonal dstrbuton of the outcome varable gven the covarates. Ths model was then used to defne a predctor of the small area characterstc of nterest that accounts for spatal structure of the data. The M-quantle GWR small area model ntegrates the concepts of robust small area estmaton and borrowng strength over space wthn a unfed modelng framework. When the functonal form of the relatonshp between the response varable and the covarates s unknown or has a complcated functonal form, an approach based on use of a nonparametrc regresson model usng penalzed splnes can offer sgnfcant advantages compared wth one based on a lnear model. By expressng the splne coeffcents n the model as random effects, Ruppert et al. (2003) show how fttng a p-splne model s equvalent to fttng a lnear mxed model. On the bass of ths property, Opsomer et al. (2008) have recently proposed a new approach to SAE that extends the unt level nested error regresson model (Battese et al., 1988) by combnng small area random effects wth a p-splne regresson model. Prates et al. (2008) have extended ths approach to the M-quantle method for the estmaton of the small area parameters usng a nonparametrc specfcaton of the condtonal M-quantle of the response varable gven the covarates. The use of bvarate p-splne approxmatons to ft nonparametrc unt level nested error and M-quantle regresson models allows for reflectng spatal varaton n the data and then uses these nonparametrc models for small area estmaton. 1

4 In ths paper we contrast SAR, unt level nested error p-splne regresson, M-quantle GWR and M-quantle splne models n terms of ther performance usng data wth geographcal nformaton. We also examne whether estmaton for out of sample areas.e. areas wth zero sample szes can be mproved usng models that borrow strength over space. The structure of the paper s as follows. In secton 2 we revew unt level mxed models wth ndependent and spatally correlated random area effects for small area estmaton. In secton 3 we present M-quantle and M-quantle GWR models for small area estmaton. In secton 4 we descrbe the nonparametrc unt level nested error and M-quantle regresson models. Estmaton of the mean squared error of the resultng small area predctors under modelng approaches s dscussed. For explorng the research questons of ths paper we employ a real survey datasets: data from the U.S. Envronmental Protecton Agency s Envronmental Montorng and Assessment Program (EMAP). The dataset contans geo-referenced and n secton 5 we use desgn-based smulaton experments for assessng the performance of the dfferent small area predctors and ther assocated estmates of mean squared error consdered n ths paper. Fnally, n secton 6 we summarze our man fndngs. 2 Unt level mxed models for small area estmaton Let x j denote a vector of p auxlary varables for each populaton unt j n small area and assume that nformaton for the varable of nterest y s avalable only from the sample. The target s to use the data to estmate varous area-specfc quanttes. A popular approach for ths purpose s to use mxed effects models wth random area effects. A lnear mxed effects model s y j = x T jβ + z j γ + ɛ j, j = 1..., n = 1,..., d (1) where β s the p 1 vector of regresson coeffcents, γ denotes a random area effect that characterzes dfferences n the condtonal dstrbuton of y gven x between the d small areas, z j s a constant whose value s known for all unts n the populaton and ɛ j s the error term assocated wth the j-th unt wthn the -th area. Conventonally, γ and ɛ j are assumed to be ndependent and normally dstrbuted wth mean zero and varances σγ 2 and σɛ 2 respectvely. The Emprcal Best Lnear Unbased Predctor (EBLUP) of the mean for small area (Battese et al., 1988; Rao, 2003) s then ˆm MX [ y j + j s j r ŷ j where ŷ j = x T j ˆβ + z j ˆγ, s denotes the n sampled unts n area, r denotes the remanng N n unts n the area and ˆβ and ˆγ are obtaned by substtutng an optmal estmate of the covarance matrx of the random effects n (1) nto the best lnear unbased estmator of β and the best lnear unbased predctor of γ respectvely. Model (1) can be extended to allow for correlated random area effects. Let the devatons v from the fxed part of the model X T β be the result of an autoregressve process wth parameter ρ and proxmty matrx W (Anseln, 1988; Cresse, 1993), then ] v = ρwv + γ v = (I ρw) 1 γ (3) where I s a d d dentty matrx. Combnng (1) and (3), wth ε ndependent of v, the model wth spatally correlated errors can be expressed as y = X T β + Z(I ρw) 1 γ + ε. (4) The error term v then has a d d Smultaneously Autoregressve (SAR) dsperson matrx gven by G = σ 2 γ [ (I ρw T )(I ρw)] 1. (5) The W matrx descrbes the neghbourhood structure of the small areas and ρ defnes the strength of the spatal relatonshp between the random effects of neghbourng areas. Under (4), the Spatal Best Lnear Unbased Predctor (Spatal BLUP) of the small area mean and ts emprcal verson (SEBLUP) are obtaned followng Henderson (1975). In partcular, the SEBLUP of the small area mean, m, s ˆm MX/SAR [ y j + j s j r ŷ j 2.1 MSE estmaton for small area estmates under the SAR mxed model An expresson for the mean squared error (MSE) under the SEBLUP model and ts estmator are obtaned followng the results of Kackar & Harvlle (1984), Prasad & Rao (1990) and Datta & Lahr (2000). More specfcally, the MSE estmator conssts of three components denoted by g 1, g 2 and g 3 : MSE[ ˆm MX/SAR ] = g 1 (σγ, 2 σε, 2 ρ) + g 2 (σγ, 2 σε, 2 ρ) + g 3 (σγ, 2 σε, 2 ρ). ] (2) (6) 2

5 The components of the MSE are due to the varablty assocated wth the estmaton of the random effects (g 1 ), the estmaton of β (g 2 ) and the estmaton of (σγ, 2 σε, 2 ρ) (g 3 ). Note that due to the ntroducton of the addtonal parameter ρ, component g 3 of the MSE s not the same as n the case of the EBLUP (Prasad & Rao, 1990). In practcal applcatons the predctor ˆm MX/SAR has to be assocated wth an estmator of MSE[ ˆm MX/SAR ]. Followng the results of Harvlle & Jeske (1992) and Zmmerman & Cresse (1992) an approxmately unbased mean squared error estmator of s gven by mse[ ˆm MX/SAR ] g 1 (ˆσ 2 γ, ˆσ 2 ε, ˆρ) + g 2 (ˆσ 2 γ, ˆσ 2 ε, ˆρ) + 2g 3, (ˆσ 2 γ, ˆσ 2 ε, ˆρ) (7) when ˆσ 2 γ, ˆσ 2 ε, ˆρ are REML estmators. See Sngh et al. (2005) and Prates & Salvat (2008) for detals. 3 M-quantle models for small area estmaton A recently proposed approach to small area estmaton s based on the use of M-quantle models (Chambers & Tzavds, 2006; Tzavds et al., 2008). A lnear M-quantle regresson model s one where the q th M-quantle Q q (x; ψ) of the condtonal dstrbuton of y gven x satsfes Q q (X; ψ) = X T β ψ (q). (8) Here ψ denotes the nfluence functon assocated wth the M-quantle. For specfed q and contnuous ψ, an estmate ˆβ ψ (q) of β ψ (q) s obtaned va teratve weghted least squares. Followng Chambers & Tzavds (2006) an alternatve to random effects for characterzng the varablty across the populaton s to use the M-quantle coeffcents of the populaton unts. For unt j wth values y j and x j, ths coeffcent s the value θ j such that Q θj (x j ; ψ) = y j. These authors observed that f a herarchcal structure does explan part of the varablty n the populaton data, unts wthn clusters (areas) defned by ths herarchy are expected to have smlar M-quantle coeffcents. When the condtonal M-quantles are assumed to follow a lnear model, wth β ψ (q) a suffcently smooth functon of q, ths suggests a predctor of m of the form ˆm MQ [ y j + j s {x T ˆβ ] j ψ (ˆθ )} j r where ˆθ s an estmate of the average value of the M-quantle coeffcents of the unts n area. Typcally ths s the average of estmates of these coeffcents for sample unts n the area. These unt level coeffcents are estmated by solvng ˆQ qj (x j ; ψ) = y j for q j wth ˆQ q denotng the estmated value of (8) at q. When there s no sample n area then ˆθ = 0.5. Tzavds et al. (2008) refer to (9) as the nave M-quantle predctor and note that ths can be based. To rectfy ths problem these authors propose a bas adjusted M-quantle predctor of m that s derved as the mean functonal of the Chambers & Dunstan (1986) (CD hereafter) estmator of the dstrbuton functon and s gven by ˆm MQ/CD = + td ˆF [ (t) y j + j s {x T ˆβ j ψ (ˆθ )} + N n n j r 3.1 M-quantle geographcally weghted model for small area estmaton (9) {y j x T ˆβ ] j ψ (ˆθ )}. (10) j s SAR mxed models are global models.e. wth such models we assume that the relatonshp we are modelng holds everywhere n the study area and we allow for spatal correlaton at dfferent herarchcal levels n the error structure. One way of ncorporatng the spatal structure of the data n the M-quantle small area model s va an M-quantle GWR model (Salvat et al., 2008). Unlke SAR mxed models, M-quantle GWR are local models that allow for a spatally non-statonary process n the mean structure of the model. Gven n observatons at a set of L locatons {u l ; l = 1,..., L; L n} wth n l data values {(y jl, x jl ); j = 1,..., n l } observed at locaton u l, an M-quantle GWR model s defned as Q q (X; ψ, u) = X T β ψ (u; q) (11) where now β ψ (u; q) vares wth u as well as wth q. The M-quantle GWR s a local model for the entre condtonal dstrbuton -not just the mean- of y gven x. Estmates of β ψ (u; q) n (11) can be obtaned by solvng L n l w(u l, u) ψ q {y jl x T jlβ ψ (u; q)}x jl = 0 (12) l=1 j=1 3

6 where ψ q (t) = 2ψ(s 1 t){qi(t > 0)+(1 q)i(t 0)} and s s a sutable robust estmate of scale such as the medan absolute devaton (MAD) estmate s = medan y jl x T jl β ψ(u; q) / It s also customary to assume a Huber type nfluence functon although other nfluence functons are also possble ψ(t) = ti( c t c) + sgn(t)i( t > c). Provded c s bounded away from zero, an teratvely re-weghted least squares algorthm can then be used to solve (12), leadng to estmates of the form { 1X ˆβ ψ (u; q) = X T W (u; q)x} T W (u; q)y. (13) In (13) y s the vector of n sample y values and X s the correspondng desgn matrx of order n p of sample x values. The matrx W (u; q) s a dagonal matrx of order n wth entres correspondng to a partcular sample observaton and equal to the product of ths observaton s spatal weght, whch depends on ts dstance from locaton u, wth the weght that ths observaton has when the sample data are used to calculate the spatally statonary M-quantle estmate ˆβ ψ (q). At ths pont we should menton that the spatal weght s derved from a spatal weghtng functon whose value depends on the dstance from sample locaton u l to u such that sample observatons wth locatons close to u receve more weght than those further away. One popular approach to defnng such a weghtng functon s to use { w(u l, u) = exp 0.5(d (ul,u)/b) 2}, where d (ul,u) denotes the Eucldean dstance between u l and u and b s the bandwdth, whch can be optmally defned usng a least squares crteron (Fotherngham et al., 2002). It should be noted, however, that alternatve weghtng functons, for example the b-square functon, can also be used. Salvat et al. (2008) also proposed a reduced M-quantle GWR that combnes local ntercepts wth global slopes and s defned as Q q (X; ψ, u) = X T β ψ (q) + δ ψ (u; q). (14) Ths s ftted n two steps. At the frst step we gnore the spatal structure n the data and estmate β ψ (q) drectly va the teratve re-weghted least squares algorthm used to ft the standard lnear M-quantle regresson model (8). At the second step geographc weghtng s appled to estmate δ ψ (u; q) usng ˆδ ψ (u; q) = n 1 L l=1 n l w(u l, u) ψ q {y jl x T ˆβ jl ψ (q)}. Hereafter we refer to (11) and (14) as the MQGWR and MQGWR-LI (Local Intercepts) models respectvely. The prmary am of ths paper s to employ the MQGWR and MQGWR-LI models for estmatng the area mean m of y. Followng Chambers & Tzavds (2006) ths can be done by frst estmatng the M-quantle GWR coeffcents {q j ; j s} of the sampled populaton unts wthout reference to the small areas of nterest. A grd-based nterpolaton procedure for dong ths under (8) s descrbed n Chambers & Tzavds (2006) and can be drectly used wth the M-quantle GWR models (11) and (14). In partcular, we adapt ths approach wth M-quantle GWR models by frst defnng a fne grd of q values over the nterval (0, 1) and then use the sample data to ft (11) or (14) for each dstnct value of q on ths grd and at each sample locaton. The M-quantle GWR coeffcent for unt j wth values y j and x j at locaton u j s computed by nterpolatng over ths grd to fnd the value q j such that Q qj (x j ; ψ, u j ) = y j. Provded there are sample observatons n area, an area-specfc M-quantle GWR coeffcent, ˆθ, can be defned as the average value of the sample M-quantle GWR coeffcents n area. Followng Salvat et al. (2008), the bas-adjusted M- quantle GWR predctor of the mean n small area s MQGW R/CD ˆm [ y j + j s j r j=1 ˆQˆθ (x j ; ψ, u j ) + N n n where ˆQˆθ (x j ; ψ, u j ) s defned ether va the MQGWR model (11) or va the MQGWR-LI (14). 3.2 MSE estmaton for small area estmates under the M-quantle GWR model ] {y j ˆQˆθ (x j ; ψ, u j )}. (15) j s Mean squared error estmaton for the M-quantle GWR small area estmates s based on the Chambers et al. (2008) estmator and s also descrbed n Salvat et al. (2008). To start wth we note that (15) can be expressed as a weghted sum of the sample y-values MQGW R/CD ˆm w T y (16) 4

7 where w = N 1 + H T n jx j N n H T n jx j. (17) j r j s Here 1 s the n-vector wth j th component equal to one whenever the correspondng sample unt s n area and s zero otherwse and { 1X H j = X T W (u j ; ˆθ )X} T W (u j ; ˆθ ). Gven the lnear representaton (16), an estmator of a frst order approxmaton to the mean squared error of ths predctor can be computed followng standard methods of robust mean squared error estmaton for lnear predctors of populaton quanttes (Royall & Cumberland, 1978). Put w = (w j ) ths estmator s of the form where λ jk = v( ˆm MQGW R/CD ) = N 2 λ jk k:n k >0 j s k { } (w j 1) 2 + (n 1) 1 (N n ) I(k = ) + wjk 2 I(k ). { y j ˆQˆθk (x j ; ψ, u j )} 2 (18) 4 Nonparametrc small area models Although very useful n many stuatons, lnear mxed models depend on dstrbutonal assumptons for the random part of the model and do not easly allow for outler robust nference. In addton, the fxed part of the model may not be flexble enough to handle cases n whch the relatonshp between the varable of nterest and the covarates s more complex than that assumed by a lnear model. Opsomer et al. (2008) extend model (1) to the case n whch the small area random effects can be combned wth a smooth, non-parametrcally specfed trend. In partcular, n the smplest case y j = m(x 1j ) + z j γ + ɛ j, j = 1..., n = 1,..., d, (19) where m( ) s an unknown smooth functon of the varable x 1. The estmator of the small area mean s [ ] NP MX ˆm y j + ŷ j j s j r (20) as n (2) where ŷ j = ˆm(x 1j ) + z j ˆγ. By usng penalzed splnes as the representaton for the non-parametrc trend, Opsomer et al. (2008) express the non-parametrc small area estmaton model as a random effects model wth ˆm(x 1j ) = ˆβ 0 + ˆβ 1 x 1j + a j δ. Then the ŷ j value ncludes the splne functon, whch s treated as a random effect, and the small area random effect. The latter can be easly extended to handle bvarate smoothng and addtve modelng. These authors proposed and studed the theoretcal propertes of the mean squared error of (20). They extended the results of Prasad & Rao (1990) and Das et al. (2008) to the case of a splne-based random effect. Opsomer et al. (2008) also proposed a bootstrap estmator for the MSE, whch performs reasonably well, but s computatonally ntensve. Prates et al. (2008) have extended ths approach to the M-quantle method for the estmaton of the small area parameters usng a nonparametrc specfcaton for the condtonal M-quantles of the response varable gven the covarates. When the functonal form of the relatonshp between the q-th M-quantle and the covarates devates from the assumed one, the lnear M-quantle regresson model can lead to based estmators of the small area parameters. When the relatonshp between the q-th M-quantle and the covarates s not lnear, a p-splnes M-quantle regresson model may have sgnfcant advantages compared to the lnear M-quantle model. The small area estmator of the mean may be taken as n (9) where the unobserved value for populaton unt r j s predcted usng ŷ j = x j ˆβψ (ˆθ ) + a j ˆν ψ (ˆθ ), where ˆβ ψ (ˆθ ) and ˆν ψ (ˆθ ) are the coeffcent vectors of the parametrc and splne components, respectvely, of the ftted p- splnes M-quantle regresson functon at ˆθ. In case of p-splnes M-quantle regresson models the bas-adjusted estmator for the mean s gven by NP MQ/CD ˆm = 1 { ŷ j + N } (y j ŷ j ), (21) N n j U s where ŷ j denotes the predcted values for the populaton unts n s and n U. NP MQ/CD Followng the approach descrbed n Chambers et al. (2008), for fxed q, the ˆm n (21) can be wrtten as lnear combnaton of the observed y j. The derved weghts are treated as fxed and a plug n estmator of the mean squared error of estmator has been obtaned by usng standard methods for robust estmaton of the varance of unbased weghted lnear 5

8 estmators (Royall & Cumberland, 1978) and by followng the results due to Chambers et al. (2008). See Salvat et al. (2008a) for detals. The use of bvarate p-splne approxmatons to ft nonparametrc unt level nested error and M-quantle regresson models allows for reflectng the spatal varaton n the data and then uses these nonparametrc models for small area estmaton. In partcular, for M-quantle models, as we have just dealt wth flexble smoothng of quantles n scatterplots, we can now handle the way n whch two contnuous varables affect the quantles of the response wthout any structural assumptons: Q q (x 1, x 2, ψ) = m ψ,q (x 1, x 2 ),.e. we can deal wth bvarate smoothng. It s of central nterest n a number of applcaton areas as envronment and publc health. It has partcular relevance when geographcally referenced responses need to be converted to maps. As seen earler, p-splnes rely on a set of bass functons to handle nonlnear structures n the data. Bvarate smoothng requres bvarate bass functons; Ruppert et al. (2003) advocate the use of radal bass functons to derve Low-rank thn plate splnes. In partcular, the followng model s assumed at quantle q for unt : m ψ,q [x 1, x 2 ; β ψ (q), γ ψ (q)] = β 0ψ (q) + β 1ψ (q)x 1 + β 2ψ (q)x 2 + a ν ψ (q). (22) Here z s the -th row of the followng n K matrx A = [C( x κ k )] 1 n [C(κ k κ k )] 1/2 1 k,k 1 k K K, (23) where C(t) = t 2 log t, x = (x 1, x 2 ) and κ k, k = 1,..., K are knots. See Prates et al. (2008) for detals on ths. The choce of knots n two dmensons s more challengng than n one. One approach could be that of layng down a rectangular lattce of knots, but ths has a tendency to waste a lot of knots when the doman defned by x 1 and x 2 has an rregular shape. In one dmenson a soluton to ths ssue s that of usng quantles. However, the extenson of the noton of quantles to more than one dmenson s not straghtforward. Two solutons suggested n lterature that provde a subset of observatons ncely scattered to cover the doman are space fllng desgns (Nychka & Saltzman, 1998) and the clara algorthm (Kaufman & Rousseeuw, 1990). The frst one s based on the maxmal separaton prncple of K ponts among the unque x and s mplemented n the felds package of the R language. The second one s based on clusterng and selects K representatve objects out of n; t s mplemented n the package cluster of R. 4.1 A note on small area estmaton for out of sample areas In some stuatons we are nterested n estmatng small area characterstcs for domans (areas) wth no sample observatons. The conventonal approach to estmatng a small area characterstc, say the mean, n ths case s synthetc estmaton. Under the mxed model (1) or the SAR mxed model (4) the synthetc mean predctor for out of sample area s MX/SY NT H ˆm where U = s r. Under M-quantle model (8) the synthetc mean predctor for out of sample area s MQ/SY NT H ˆm j U x T j ˆβ, (24) j U x T j ˆβ ψ (0.5). (25) We note that wth synthetc estmaton all varaton n the area-specfc predctons comes from the area-specfc auxlary nformaton. One way of potentally mprovng the conventonal synthetc estmaton for out of sample areas s by usng a model that borrows strength over space such as an M-quantle GWR model and nonparametrc unt level nested error and M-quantle regresson models. In ths case a synthetc-type mean predctors for out of sample area are defned by MQGW R/SY NT H ˆm NP MQ/SY NT H ˆm NP MX/SY NT H ˆm j U ˆQ0.5 (x j ; ψ, u j ) (26) j U ˆm(x 1j, u j ) (27) j U [x j ˆβψ (0.5) + z j ˆν ψ (0.5, u j )]. (28) Emprcal results that address the ssue of out of sample area estmaton are set out n sectons 5. 6

9 5 Desgn-based smulaton study In ths secton we present results from a smulaton study that was used to examne the performance of the small area predctors dscussed n the prevous sectons. We present a desgn-based smulaton usng data from the Envronmental Montorng and Assessment Program (EMAP) that forms part of the Space Tme Aquatc Resources Modellng and Analyss Program (STARMAP) at Colorado State Unversty. The survey data used n ths desgn-based smulaton comes from the U.S. Envronmental Protecton Agency s Envronmental Montorng and Assessment Program (EMAP) Northeast lakes survey (Larsen et al., 1997). Between 1991 and 1995, researchers from the U.S. Envronmental Protecton Agency (EPA) conducted an envronmental health study of the lakes n the north-eastern states of the U.S. For ths study, a sample of 334 lakes was selected from the populaton of 21, 026 lakes n these states. The lakes formng ths populaton were grouped accordng to dgt Hydrologc Unt Codes (HUCs) of whch 64 contaned less than 5 observatons and 27 dd not have any observatons. The varable of nterest was Acd Neutralzng Capacty (ANC), an ndcator of the acdfcaton rsk of water bodes. Snce some lakes were vsted several tmes durng the study perod and some of these were measured at more than one ste, the total number of observed stes was 349 wth a total of 551 measurements. In addton to ANC values, the EMAP data set also contaned the elevaton and geographcal coordnates of the centrod of each lake n the target area (HUC). For sampled locatons we know the exact spatal coordnates of the correspondng locaton. For non-sampled locatons the centrod of the lake s known. Hence, detaled nformaton on the spatal coordnates for non-sampled locatons exsts as the geography defned by the lakes s below the geography of nterest defned by the HUCs. The am of ths desgn-based smulaton was (a) to compare the performance of the dfferent small area predctors of the mean of ANC n each HUC and (b) to evaluate the performance of the dfferent predctors for estmatng the mean ANC for out of sample HUCs. In order to do ths, we frst created a populaton of ANC values wth smlar spatal characterstcs to that of the lakes sampled by EMAP. A total of 200 ndependent random samples were then taken from each HUC, that had been sampled by EMAP, wth sample szes set equal to the max(5, n ) where n s the sample sze of each HUC n the orgnal EMAP dataset. No observatons were taken from HUCs that had not been sampled by EMAP. Ths process led to a total sample sze of 652 ANC values from 86 HUCs. In order to generate a populaton dataset that had smlar spatal structure to that of the EMAP sample data, we allocated ANC values to the non-sampled lakes as follows: (1) we frst randomly ordered the non-sampled locatons n order to avod lst order bas and gave each sampled locaton a donor weght equal to the nteger component of ts survey weght mnus 1; (2) takng each non-sample locaton n turn, we chose a sample locaton as a donor for the j th non-sample locaton by selectng one of the ANC values of the EMAP sample locatons wth probablty proportonal to w(u j, u) = exp{ 0.5(d (uj,u)/b) 2 }. Here d (uj,u) s the Eucldean dstance from the j th non-sample locaton u j to the locaton u of a sampled locaton and b s the GWR bandwdth estmated from the EMAP data; and (3) we reduced the donor weght of the selected donor locaton by 1. We compare the followng small area predctors (a) EBLUP (2), (b) M-quantle CD (10) (MQ), (c) M-quantle GWR CD (15) under model (11) (MQGWR), (d) M-quantle GWR CD (15) under the local ntercepts model (14) (MQGWR-LI), (e) SEBLUP (6) (f) nonparametrc EBLUP (20) (NPEBLUP) and (g) nonparametrc M-quantle CD (21) (NPMQ). For the M-quantle GWR predctors we use the centrod of the lake. The relatve bas (RB) and the relatve root mean squared error (RRMSE) of estmates of the mean value of ANC n each HUC were computed. Before presentng the results from ths smulaton study we would lke to show some model and spatal dagnostcs. Fgure 1 shows normal probablty plots of level 1 and level 2 resduals obtaned by fttng a two-level (level 1 s the lake and level 2 s the HUC) mxed model to the synthetc populaton data. The normal probablty plots ndcate that the Gaussan assumptons of the mxed model are not met. Hence, the use of a model that relaxes these assumptons, such as the M-quantle model, can be justfed n ths case. In order to detect whether there s spatal autocorrelaton n the EMAP data we computed the Moran s I coeffcent. The standardzed Moran s I s analogous to the correlaton coeffcent, and ts values range from 1 (strong postve spatal autocorrelaton) to -1 (strong negatve spatal autocorrelaton). For the EMAP data Moran s I = 0.61 ndcatng a postve spatal correlaton. There s also evdence of a non-statonary process. In partcular, usng an ANOVA test proposed by Brundson et al. (1999) we rejected the null hypothess of statonarty of the model parameters. Based on the spatal dagnostcs we expect that ncorporatng the spatal nformaton n small area estmaton may lead to gans n effcency. The results set out n Tables 1 and 2 show the across areas and smulatons dstrbuton of relatve bas and relatve root mean squared error for n sample and out of sample areas respectvely. Focusng frst on Table 1 we note that all small area predctors based on the dfferent varants of the M-quantle GWR model have sgnfcantly lower relatve bas than the EBLUP SEBLUP, and NPEBLUP predctors wth the MQGWR predctor performng best. Examnng the performance n terms of relatve root mean squared error we note that the small area predctors that account for the spatal structure of the data have on average smaller root mean squared errors wth the NPEBLUP, SEBLUP and MQGWR predctors performng best. The ncreased relatve root mean squared error of the MQGWR predctors can be explaned by the bas varance trade off assocated wth the use of robust methods. That s, although by usng the M-quantle GWR model we reduce the bas of the pont estmates, the MQGWR predctors have hgher varablty. One way of potentally tacklng ths problem s by makng the M-quantle GWR model less robust for example by settng n the Huber nfluence functon c > These results also show that 7

10 Sample Quantles Sample Quantles Theoretcal Quantles Theoretcal Quantles Fgure 1: Normal probablty plots of level 2 (left) and level 1 resduals (rght) derved by fttng a two level lnear mxed model to the synthetc populaton data. there s a substantal number of n-sample HUCs where the MQGWR predctor has lower RRMSE than the NPEBLUP and SEBLUP predctors. Focusng now on Table 2 we note that for out of sample areas NPMQ and MQGWR-based small area predctors have lower relatve bases and lower root mean squared errors than the EBLUP, NPEBLUP and SEBLUP predctors. Ths supports our orgnal hypothess that the M-quantle GWR model offers a straghtforward approach for mprovng synthetc estmaton for out of sample areas. The performance of the SEBLUP predctor n ths case may be surprsng. However, we should bear n mnd that for out of sample areas we use a synthetc SEBLUP. A more elaborate method for out of sample areas under the SAR model has been proposed by Sae & Chambers (2005). Fgures 2 and 3 show how the dfferent mean squared estmators track the true mean squared error of the dfferent predctors. Here we see that mean squared estmator descrbed n Tzavds et al. (2008), and ts verson (18) under the M-quantle GWR model, perform well n terms of trackng the true mean squared error of the M-quantle small area predctors. Fnally, we see that the Prasad-Rao type MSE estmators of the EBLUP, NPEBLUP and SEBLUP perform poorly n ths applcaton as far as trackng the area-specfc mean squared error s concerned. Ths phenomenon has been also reported n other desgn-based studes (Chambers et al., 2008). As the model dagnostcs have already demonstrated, for ths data the Gaussan assumptons of the mxed model are not satsfed. Ths provdes a further explanaton for the performance of the Prasad-Rao type mean squared estmators n ths case. 6 Conclusons In ths paper we contrasted dfferent approaches for borrowng strength over space n small area estmaton usng parametrc, semparametrc and nonparametrc small area models. Our results show that ncorporatng the spatal nformaton n small area estmaton can lead to sgnfcant gans n the effcency of the small area estmates. The penalzed splnes model appears to be a useful tool when the functonal form of the relatonshp between the varable of nterest and the covarates s left unspecfed and the data are characterzed by complex patterns of spatal dependency. An advantage of the M-quantle GWR-based estmators s that they perform better than the SEBLUP, NPEBLUP estmator for estmatng parameters for out of sample areas. One approach for potentally mprovng the performance of the SEBLUP estmator for out of sample areas s to use the Sae & Chambers (2005) SAR mxed model. A further advantage of the M-quantle GWR model s that t allows for outler robust nference. As we llustrated n secton 5, the volaton of the assumptons of the mxed model may lead to substantal bas n the small area estmates derved wth the EBLUP SEBLUP and NPEBLUP predctors. The volaton of the assumptons of the mxed model also affects the performance of the Prasad-Rao type mean squared error estmators. On the other hand, the use of a robust approach to small area estmaton tackles the problem of bas though at the expense of hgher varablty. Approaches to balancng ths bas varance trade off were brefly descrbed n secton 5. In a recent paper Snha & Rao (2008) proposed the use of a robust mxed model for small area estmaton. The robust mxed model can be drectly compared to the M-quantle 8

11 Table 1: Desgn-based smulaton results usng the EMAP data. Results show the dstrbuton of Relatve Bas (RB) and Relatve Root Mean Squared Error (RRMSE) over areas and smulatons for the 86 sampled HUCs. Percentle of across area dstrbuton Predctor Mean Relatve Bas (%) EBLUP SEBLUP MQ MQGWR MQGWR-LI NPEBLUP NPMQ RRMSE (%) EBLUP SEBLUP MQ MQGWR MQGWR-LI NPEBLUP NPMQ Table 2: Desgn-based smulaton results usng the EMAP data. Results show the dstrbuton of Relatve Bas (RB) and Relatve Root Mean Squared Error (RRMSE) over areas and smulatons for the 27 out of sample HUCs. Percentle of across area dstrbuton Predctor Mean Relatve Bas (%) EBLUP SEBLUP MQ MQGWR MQGWR-LI NPEBLUP NPMQ RRMSE (%) EBLUP SEBLUP MQ MQGWR MQGWR-LI NPEBLUP NPMQ

12 RMSE RMSE HUC HUC RMSE RMSE HUC HUC Fgure 2: HUC-specfc values of actual desgn-based RMSE (sold lne) and average estmated RMSE (dashed lne). Top left s the approxmaton to the RMSE of the MQ predctor. Top rght s the approxmaton to the RMSE of the MQGWR predctor. Bottom left s the approxmaton to the RMSE of the MQGWR-LI predctor and bottom rght s the approxmaton to the RMSE of the NPMQ (ag.sp) predctor. Estmates of the RMSE under the dfferent models are obtaned usng the mean squared error estmator suggested by Chambers et al. (2008). 10

13 RMSE RMSE HUC HUC RMSE HUC Fgure 3: HUC-specfc values of actual desgn-based RMSE (sold lne) and average estmated RMSE (dashed lne). Top left s the Prasad & Rao (1990) approxmaton to the RMSE of the EBLUP predctor. Top rght s the approxmaton to the RMSE of the SEBLUP predctor usng the RMSE estmator of secton 2.1. Bottom left s the approxmaton to the RMSE of the NPEBLUPusng the RMSE estmator suggested by Opsomer et al. (2008). 11

14 small area model and we currently workng on comparng the two models. Extendng further the outler robust mxed model nto an outler robust SAR mxed model wll further mprove the collecton of small area estmaton tools. References ANSELIN, L. (1988). Spatal Econometrcs. Methods and Models. Boston: Kluwer Academc Publshers. BANERJEE, S., CARLIN, B. & GELFAND, A. (2004). Herarchcal Modelng and Analyss for Spatal Data. New York: Chapman and Hall. BATTESE, G. E., HARTER, R.M. & FULLER, W.A.(1988). An error-components model for predcton of county crop areas usng survey and satellte data. Journal of the Amercan Statstcal Assocaton 83, BRUNDSON, C., FOTHERINGHAM, A.S. & CHARLTON, M.(1996). Geographcally weghted regresson: a method for explorng spatal nonstatonarty. Geographcal Analyss 28, BRUNDSON, C., FOTHERINGHAM, A.S. & CHARLTON, M.(1999). Some notes on parametrc sgnfcance tests for geographcally weghted regresson. Journal of Regonal Scence 39, CHAMBERS, R. & DUNSTAN, R. (1986). Estmatng dstrbuton functon from survey data. Bometrka 73, CHAMBERS, R. & TZAVIDIS, N. (2006). M-quantle Models for Small Area Estmaton. Bometrka 93, CHAMBERS, R., TZAVIDIS, N. & CHANDRA, H. (2008). On robust mean squared error estmaton for lnear predctors for domans. [paper submtted for publcaton, avalable upon request] CRESSIE, N. (1993). Statstcs for Spatal Data. New York: John Wley & Sons. DAS, K., JIANG, J. & RAO, J.N.K. (2004). Mean squared error of emprcal predctor. Ann. Statst. 32, DATTA, G.S. & LAHIRI, P. (2000). A Unfed Measure of Uncertanty of Estmates for Best Lnear Unbased Predctors n Small Area Estmaton Problem. Statstca Snca 10, FOTHERINGHAM, A.S., BRUNDSON, C. & CHARLTON, M.(1996). Two technques for explorng non-statonarty n geographcal data. Geographcal Systems 4, FOTHERINGHAM, A.S., BRUNDSON, C. & CHARLTON, M.(1996). Geographcally Weghted Regresson West Sussex: John Wley & Sons. KACKAR, R. & HARVILLE, D. (1984). Approxmatons for standard errors of estmators for fxed and random effects n mxed models. Journal of the Amercan Statstcal Assocaton 79, KAUFMAN, L. & ROUSSEEUW, P. (1990). Fndng Groups n Data: An Introducton to Cluster Analyss. New York: Wley. HENDERSON, C. (1975). Best lnear unbased estmaton and predcton under a selecton model. Bometrcs 31, HARVILLE, D.A. & JESKE, D.R. (1992). Mean squared error of estmaton or predcton under a general lnear model. Journal of the Amercan Statstcal Assocaton 87, LARSEN, D. P., KINCAID, T. M., JACOBS, S. E. & URQUHART, N. S.(2001). Desgns for evaluatng local and regonal scale trends. Boscence 51, LONGFORD, N.T. (2007). On standard errors of model-based small area estmators. Survey Methodology 33, NYCHKA, D. & SALTZMAN, N. (1998). Desgn of ar qualty montorng networks. In Nychka, Douglas, Pegorsch, Walter W. and Cox, Lawrence H. (eds), Case studes n envronmental statstcs OPSOMER, J. D. CLAESKENS, G., RANALLI, M. G., KAUERMANN, G. & BREIDT, F. J.(2008). Nonparametrc small area estmaton usng penalzed splne regresson. Royal Statstcal Socety, Seres B 70, PETRUCCI, A. & SALVATI, N. (2006). Small area estmaton for spatal correlaton n watershed eroson assessment. Journal of Agrcultural, Bologcal and Envronmental Statstcs 11, PRASAD, N. & RAO, J. (1990). The estmaton of the mean squared error of small-area estmators. Journal of the Amercan Statstcal Assocaton 85,

15 PRATESI, M. & SALVATI, N. (2007). Small area estmaton: the EBLUP estmator based on spatally correlated random area effects. Statstcal Methods & Applcatons 17, PRATESI, M., RANALLI, M. G.& SALVATI, N. (2008). Semparametrc M-quantle regresson for estmatng the proporton of acdc lakes n 8-dgt HUCs of the Northeastern US. Envronmetrcs 19, RAO, J.N.K., KOVAR, J.G. & MANTEL, H.J. (1990). On Estmatng Dstrbuton Functons and Quantles from Survey Data Usng Auxlary Informaton. Bometrka 77, RAO, J. N. K. (2003). Small Area Estmaton. London: Wley. ROYALL, R.M. & CUMBERLAND, W.G. (1978). Varance estmaton n fnte populaton samplng. Journal of the Amercan Statstcal Assocaton 73, RUPPERT, D., WAND, M. P.& CARROLL, R.(2003). Semparametrc Regresson. Cambrdge: Cam- brdge Unversty Press. SAEI, A. & CHAMBERS, R. (2005). Emprcal Best Lnear Unbased Predcton for Out of Sample Areas. Workng Paper M05/03, Southampton Statstcal Scences Research Insttute, Unversty of Southampton. SALVATI, N., TZAVIDIS, N., PRATESI, M. & CHAMBERS, R. (2008). Small Area Estmaton Va M-quantle Geographcally Weghted Regresson. [paper submtted for publcaton, avalable upon request] SALVATI, N., RANALLI, M.G. & PRATESI, M. (2008a). Nonparametrc M-quantle Regresson usng Penalzed Splnes n Small Area Estmaton [paper submtted for publcaton, avalable upon request] SINGH, B., SHUKLA, G. & KUNDU, D. (2005). Spato-temporal models n small area estmaton. Survey Methodology 31, SINHA, S.K. & RAO, J.N.K. (2008). Robust Small Area Estmaton. [paper submtted for publcaton] TZAVIDIS, N., MARCHETTI, S. & CHAMBERS, R. (2008). Robust predcton of small area means and dstrbutons. [paper submtted for publcaton, avalable upon request] YU, D.L. & WU, C. (2004). Understandng populaton segregaton from Landsat ETM+magery: a geographcally weghted regresson approach. GISence and Remote Sensng 41, ZIMMERMAN, D.L. & CRESSIE, N. (1992). Mean squared predcton error n the spatal lnear model wth estmated covarance parameters. Ann. Inst. Stat. Math. 44,

Small Area Estimation Under Spatial Nonstationarity

Small Area Estimation Under Spatial Nonstationarity Small Area Estmaton Under Spatal Nonstatonarty Hukum Chandra Indan Agrcultural Statstcs Research Insttute, New Delh Ncola Salvat Unversty of Psa Ray Chambers Unversty of Wollongong Nkos Tzavds Unversty

More information

On Outlier Robust Small Area Mean Estimate Based on Prediction of Empirical Distribution Function

On Outlier Robust Small Area Mean Estimate Based on Prediction of Empirical Distribution Function On Outler Robust Small Area Mean Estmate Based on Predcton of Emprcal Dstrbuton Functon Payam Mokhtaran Natonal Insttute of Appled Statstcs Research Australa Unversty of Wollongong Small Area Estmaton

More information

Bias-correction under a semi-parametric model for small area estimation

Bias-correction under a semi-parametric model for small area estimation Bas-correcton under a sem-parametrc model for small area estmaton Laura Dumtrescu, Vctora Unversty of Wellngton jont work wth J. N. K. Rao, Carleton Unversty ICORS 2017 Workshop on Robust Inference for

More information

Statistics for Economics & Business

Statistics for Economics & Business Statstcs for Economcs & Busness Smple Lnear Regresson Learnng Objectves In ths chapter, you learn: How to use regresson analyss to predct the value of a dependent varable based on an ndependent varable

More information

Department of Quantitative Methods & Information Systems. Time Series and Their Components QMIS 320. Chapter 6

Department of Quantitative Methods & Information Systems. Time Series and Their Components QMIS 320. Chapter 6 Department of Quanttatve Methods & Informaton Systems Tme Seres and Ther Components QMIS 30 Chapter 6 Fall 00 Dr. Mohammad Zanal These sldes were modfed from ther orgnal source for educatonal purpose only.

More information

Semiparametric geographically weighted generalised linear modelling in GWR 4.0

Semiparametric geographically weighted generalised linear modelling in GWR 4.0 Semparametrc geographcally weghted generalsed lnear modellng n GWR 4.0 T. Nakaya 1, A. S. Fotherngham 2, M. Charlton 2, C. Brunsdon 3 1 Department of Geography, Rtsumekan Unversty, 56-1 Tojn-kta-mach,

More information

Comparison of Regression Lines

Comparison of Regression Lines STATGRAPHICS Rev. 9/13/2013 Comparson of Regresson Lnes Summary... 1 Data Input... 3 Analyss Summary... 4 Plot of Ftted Model... 6 Condtonal Sums of Squares... 6 Analyss Optons... 7 Forecasts... 8 Confdence

More information

Econ107 Applied Econometrics Topic 3: Classical Model (Studenmund, Chapter 4)

Econ107 Applied Econometrics Topic 3: Classical Model (Studenmund, Chapter 4) I. Classcal Assumptons Econ7 Appled Econometrcs Topc 3: Classcal Model (Studenmund, Chapter 4) We have defned OLS and studed some algebrac propertes of OLS. In ths topc we wll study statstcal propertes

More information

Chapter 11: Simple Linear Regression and Correlation

Chapter 11: Simple Linear Regression and Correlation Chapter 11: Smple Lnear Regresson and Correlaton 11-1 Emprcal Models 11-2 Smple Lnear Regresson 11-3 Propertes of the Least Squares Estmators 11-4 Hypothess Test n Smple Lnear Regresson 11-4.1 Use of t-tests

More information

A Robust Method for Calculating the Correlation Coefficient

A Robust Method for Calculating the Correlation Coefficient A Robust Method for Calculatng the Correlaton Coeffcent E.B. Nven and C. V. Deutsch Relatonshps between prmary and secondary data are frequently quantfed usng the correlaton coeffcent; however, the tradtonal

More information

/ n ) are compared. The logic is: if the two

/ n ) are compared. The logic is: if the two STAT C141, Sprng 2005 Lecture 13 Two sample tests One sample tests: examples of goodness of ft tests, where we are testng whether our data supports predctons. Two sample tests: called as tests of ndependence

More information

Basic Business Statistics, 10/e

Basic Business Statistics, 10/e Chapter 13 13-1 Basc Busness Statstcs 11 th Edton Chapter 13 Smple Lnear Regresson Basc Busness Statstcs, 11e 009 Prentce-Hall, Inc. Chap 13-1 Learnng Objectves In ths chapter, you learn: How to use regresson

More information

Computation of Higher Order Moments from Two Multinomial Overdispersion Likelihood Models

Computation of Higher Order Moments from Two Multinomial Overdispersion Likelihood Models Computaton of Hgher Order Moments from Two Multnomal Overdsperson Lkelhood Models BY J. T. NEWCOMER, N. K. NEERCHAL Department of Mathematcs and Statstcs, Unversty of Maryland, Baltmore County, Baltmore,

More information

Uncertainty as the Overlap of Alternate Conditional Distributions

Uncertainty as the Overlap of Alternate Conditional Distributions Uncertanty as the Overlap of Alternate Condtonal Dstrbutons Olena Babak and Clayton V. Deutsch Centre for Computatonal Geostatstcs Department of Cvl & Envronmental Engneerng Unversty of Alberta An mportant

More information

Negative Binomial Regression

Negative Binomial Regression STATGRAPHICS Rev. 9/16/2013 Negatve Bnomal Regresson Summary... 1 Data Input... 3 Statstcal Model... 3 Analyss Summary... 4 Analyss Optons... 7 Plot of Ftted Model... 8 Observed Versus Predcted... 10 Predctons...

More information

4 Analysis of Variance (ANOVA) 5 ANOVA. 5.1 Introduction. 5.2 Fixed Effects ANOVA

4 Analysis of Variance (ANOVA) 5 ANOVA. 5.1 Introduction. 5.2 Fixed Effects ANOVA 4 Analyss of Varance (ANOVA) 5 ANOVA 51 Introducton ANOVA ANOVA s a way to estmate and test the means of multple populatons We wll start wth one-way ANOVA If the populatons ncluded n the study are selected

More information

Statistics for Managers Using Microsoft Excel/SPSS Chapter 13 The Simple Linear Regression Model and Correlation

Statistics for Managers Using Microsoft Excel/SPSS Chapter 13 The Simple Linear Regression Model and Correlation Statstcs for Managers Usng Mcrosoft Excel/SPSS Chapter 13 The Smple Lnear Regresson Model and Correlaton 1999 Prentce-Hall, Inc. Chap. 13-1 Chapter Topcs Types of Regresson Models Determnng the Smple Lnear

More information

Chapter 5 Multilevel Models

Chapter 5 Multilevel Models Chapter 5 Multlevel Models 5.1 Cross-sectonal multlevel models 5.1.1 Two-level models 5.1.2 Multple level models 5.1.3 Multple level modelng n other felds 5.2 Longtudnal multlevel models 5.2.1 Two-level

More information

Outlier Robust Small Area Estimation

Outlier Robust Small Area Estimation Unversty of Wollongong Research Onlne Centre for Statstcal & Survey Methodology Workng Paper Seres Faculty of Engneerng and Informaton Scences 009 Outler Robust Small Area Estmaton R. Chambers Unversty

More information

Durban Watson for Testing the Lack-of-Fit of Polynomial Regression Models without Replications

Durban Watson for Testing the Lack-of-Fit of Polynomial Regression Models without Replications Durban Watson for Testng the Lack-of-Ft of Polynomal Regresson Models wthout Replcatons Ruba A. Alyaf, Maha A. Omar, Abdullah A. Al-Shha ralyaf@ksu.edu.sa, maomar@ksu.edu.sa, aalshha@ksu.edu.sa Department

More information

Spatial Modelling of Peak Frequencies of Brain Signals

Spatial Modelling of Peak Frequencies of Brain Signals Malaysan Journal of Mathematcal Scences 3(1): 13-6 (9) Spatal Modellng of Peak Frequences of Bran Sgnals 1 Mahendran Shtan, Hernando Ombao, 1 Kok We Lng 1 Department of Mathematcs, Faculty of Scence, and

More information

Comparison of the Population Variance Estimators. of 2-Parameter Exponential Distribution Based on. Multiple Criteria Decision Making Method

Comparison of the Population Variance Estimators. of 2-Parameter Exponential Distribution Based on. Multiple Criteria Decision Making Method Appled Mathematcal Scences, Vol. 7, 0, no. 47, 07-0 HIARI Ltd, www.m-hkar.com Comparson of the Populaton Varance Estmators of -Parameter Exponental Dstrbuton Based on Multple Crtera Decson Makng Method

More information

1. Inference on Regression Parameters a. Finding Mean, s.d and covariance amongst estimates. 2. Confidence Intervals and Working Hotelling Bands

1. Inference on Regression Parameters a. Finding Mean, s.d and covariance amongst estimates. 2. Confidence Intervals and Working Hotelling Bands Content. Inference on Regresson Parameters a. Fndng Mean, s.d and covarance amongst estmates.. Confdence Intervals and Workng Hotellng Bands 3. Cochran s Theorem 4. General Lnear Testng 5. Measures of

More information

Model Based Direct Estimation of Small Area Distributions

Model Based Direct Estimation of Small Area Distributions Unversty of Wollongong Research Onlne Centre for Statstcal & Survey Methodology Workng Paper Seres Faculty of Engneerng and Informaton Scences 2010 Model Based Drect Estmaton of Small Area Dstrbutons Ncola

More information

Psychology 282 Lecture #24 Outline Regression Diagnostics: Outliers

Psychology 282 Lecture #24 Outline Regression Diagnostics: Outliers Psychology 282 Lecture #24 Outlne Regresson Dagnostcs: Outlers In an earler lecture we studed the statstcal assumptons underlyng the regresson model, ncludng the followng ponts: Formal statement of assumptons.

More information

Lecture 6: Introduction to Linear Regression

Lecture 6: Introduction to Linear Regression Lecture 6: Introducton to Lnear Regresson An Manchakul amancha@jhsph.edu 24 Aprl 27 Lnear regresson: man dea Lnear regresson can be used to study an outcome as a lnear functon of a predctor Example: 6

More information

The Multiple Classical Linear Regression Model (CLRM): Specification and Assumptions. 1. Introduction

The Multiple Classical Linear Regression Model (CLRM): Specification and Assumptions. 1. Introduction ECONOMICS 5* -- NOTE (Summary) ECON 5* -- NOTE The Multple Classcal Lnear Regresson Model (CLRM): Specfcaton and Assumptons. Introducton CLRM stands for the Classcal Lnear Regresson Model. The CLRM s also

More information

x i1 =1 for all i (the constant ).

x i1 =1 for all i (the constant ). Chapter 5 The Multple Regresson Model Consder an economc model where the dependent varable s a functon of K explanatory varables. The economc model has the form: y = f ( x,x,..., ) xk Approxmate ths by

More information

STAT 3008 Applied Regression Analysis

STAT 3008 Applied Regression Analysis STAT 3008 Appled Regresson Analyss Tutoral : Smple Lnear Regresson LAI Chun He Department of Statstcs, The Chnese Unversty of Hong Kong 1 Model Assumpton To quantfy the relatonshp between two factors,

More information

BOOTSTRAP METHOD FOR TESTING OF EQUALITY OF SEVERAL MEANS. M. Krishna Reddy, B. Naveen Kumar and Y. Ramu

BOOTSTRAP METHOD FOR TESTING OF EQUALITY OF SEVERAL MEANS. M. Krishna Reddy, B. Naveen Kumar and Y. Ramu BOOTSTRAP METHOD FOR TESTING OF EQUALITY OF SEVERAL MEANS M. Krshna Reddy, B. Naveen Kumar and Y. Ramu Department of Statstcs, Osmana Unversty, Hyderabad -500 007, Inda. nanbyrozu@gmal.com, ramu0@gmal.com

More information

4.3 Poisson Regression

4.3 Poisson Regression of teratvely reweghted least squares regressons (the IRLS algorthm). We do wthout gvng further detals, but nstead focus on the practcal applcaton. > glm(survval~log(weght)+age, famly="bnomal", data=baby)

More information

2016 Wiley. Study Session 2: Ethical and Professional Standards Application

2016 Wiley. Study Session 2: Ethical and Professional Standards Application 6 Wley Study Sesson : Ethcal and Professonal Standards Applcaton LESSON : CORRECTION ANALYSIS Readng 9: Correlaton and Regresson LOS 9a: Calculate and nterpret a sample covarance and a sample correlaton

More information

Statistics for Business and Economics

Statistics for Business and Economics Statstcs for Busness and Economcs Chapter 11 Smple Regresson Copyrght 010 Pearson Educaton, Inc. Publshng as Prentce Hall Ch. 11-1 11.1 Overvew of Lnear Models n An equaton can be ft to show the best lnear

More information

Nonstationary Fay-Herriot Model for Small Area Estimation

Nonstationary Fay-Herriot Model for Small Area Estimation Nonstatonary Fay-Herrot Model for Small Area Estmaton Hkm Chandra Indan Agrcltral Statstcs Research Insttte, New Delh E-mal: hchandra@asr.res.n Ncola Salvat Unversty of Psa Ray Chambers Unversty of Wollongong

More information

Correlation and Regression. Correlation 9.1. Correlation. Chapter 9

Correlation and Regression. Correlation 9.1. Correlation. Chapter 9 Chapter 9 Correlaton and Regresson 9. Correlaton Correlaton A correlaton s a relatonshp between two varables. The data can be represented b the ordered pars (, ) where s the ndependent (or eplanator) varable,

More information

Department of Statistics University of Toronto STA305H1S / 1004 HS Design and Analysis of Experiments Term Test - Winter Solution

Department of Statistics University of Toronto STA305H1S / 1004 HS Design and Analysis of Experiments Term Test - Winter Solution Department of Statstcs Unversty of Toronto STA35HS / HS Desgn and Analyss of Experments Term Test - Wnter - Soluton February, Last Name: Frst Name: Student Number: Instructons: Tme: hours. Ads: a non-programmable

More information

Robust Small Area Estimation Using a Mixture Model

Robust Small Area Estimation Using a Mixture Model Robust Small Area Estmaton Usng a Mxture Model Jule Gershunskaya U.S. Bureau of Labor Statstcs Partha Lahr JPSM, Unversty of Maryland, College Park, USA ISI Meetng, Dubln, August 23, 2011 Parameter of

More information

STAT 511 FINAL EXAM NAME Spring 2001

STAT 511 FINAL EXAM NAME Spring 2001 STAT 5 FINAL EXAM NAME Sprng Instructons: Ths s a closed book exam. No notes or books are allowed. ou may use a calculator but you are not allowed to store notes or formulas n the calculator. Please wrte

More information

Estimation: Part 2. Chapter GREG estimation

Estimation: Part 2. Chapter GREG estimation Chapter 9 Estmaton: Part 2 9. GREG estmaton In Chapter 8, we have seen that the regresson estmator s an effcent estmator when there s a lnear relatonshp between y and x. In ths chapter, we generalzed the

More information

Statistics II Final Exam 26/6/18

Statistics II Final Exam 26/6/18 Statstcs II Fnal Exam 26/6/18 Academc Year 2017/18 Solutons Exam duraton: 2 h 30 mn 1. (3 ponts) A town hall s conductng a study to determne the amount of leftover food produced by the restaurants n the

More information

Lecture 3 Stat102, Spring 2007

Lecture 3 Stat102, Spring 2007 Lecture 3 Stat0, Sprng 007 Chapter 3. 3.: Introducton to regresson analyss Lnear regresson as a descrptve technque The least-squares equatons Chapter 3.3 Samplng dstrbuton of b 0, b. Contnued n net lecture

More information

Spatial Statistics and Analysis Methods (for GEOG 104 class).

Spatial Statistics and Analysis Methods (for GEOG 104 class). Spatal Statstcs and Analyss Methods (for GEOG 104 class). Provded by Dr. An L, San Dego State Unversty. 1 Ponts Types of spatal data Pont pattern analyss (PPA; such as nearest neghbor dstance, quadrat

More information

18. SIMPLE LINEAR REGRESSION III

18. SIMPLE LINEAR REGRESSION III 8. SIMPLE LINEAR REGRESSION III US Domestc Beers: Calores vs. % Alcohol Ftted Values and Resduals To each observed x, there corresponds a y-value on the ftted lne, y ˆ ˆ = α + x. The are called ftted values.

More information

Chapter 2 - The Simple Linear Regression Model S =0. e i is a random error. S β2 β. This is a minimization problem. Solution is a calculus exercise.

Chapter 2 - The Simple Linear Regression Model S =0. e i is a random error. S β2 β. This is a minimization problem. Solution is a calculus exercise. Chapter - The Smple Lnear Regresson Model The lnear regresson equaton s: where y + = β + β e for =,..., y and are observable varables e s a random error How can an estmaton rule be constructed for the

More information

Here is the rationale: If X and y have a strong positive relationship to one another, then ( x x) will tend to be positive when ( y y)

Here is the rationale: If X and y have a strong positive relationship to one another, then ( x x) will tend to be positive when ( y y) Secton 1.5 Correlaton In the prevous sectons, we looked at regresson and the value r was a measurement of how much of the varaton n y can be attrbuted to the lnear relatonshp between y and x. In ths secton,

More information

Dr. Shalabh Department of Mathematics and Statistics Indian Institute of Technology Kanpur

Dr. Shalabh Department of Mathematics and Statistics Indian Institute of Technology Kanpur Analyss of Varance and Desgn of Experment-I MODULE VII LECTURE - 3 ANALYSIS OF COVARIANCE Dr Shalabh Department of Mathematcs and Statstcs Indan Insttute of Technology Kanpur Any scentfc experment s performed

More information

x = , so that calculated

x = , so that calculated Stat 4, secton Sngle Factor ANOVA notes by Tm Plachowsk n chapter 8 we conducted hypothess tests n whch we compared a sngle sample s mean or proporton to some hypotheszed value Chapter 9 expanded ths to

More information

28. SIMPLE LINEAR REGRESSION III

28. SIMPLE LINEAR REGRESSION III 8. SIMPLE LINEAR REGRESSION III Ftted Values and Resduals US Domestc Beers: Calores vs. % Alcohol To each observed x, there corresponds a y-value on the ftted lne, y ˆ = βˆ + βˆ x. The are called ftted

More information

Factor models with many assets: strong factors, weak factors, and the two-pass procedure

Factor models with many assets: strong factors, weak factors, and the two-pass procedure Factor models wth many assets: strong factors, weak factors, and the two-pass procedure Stanslav Anatolyev 1 Anna Mkusheva 2 1 CERGE-EI and NES 2 MIT December 2017 Stanslav Anatolyev and Anna Mkusheva

More information

University, Bogor, Indonesia.

University, Bogor, Indonesia. ROBUST SMALL AREA ESTIMATION FOR HOUSEHOLD CONSUMPTION EXPENDITURE QUANTILES USING M-QUANTILE APPROACH (CASE STUDY: POVERTY INDICATOR DATA IN BOGOR DISTRICT) Kusman Sadk 1,a), Grnoto 1,b), Indahwat 1,c)

More information

Chapter 13: Multiple Regression

Chapter 13: Multiple Regression Chapter 13: Multple Regresson 13.1 Developng the multple-regresson Model The general model can be descrbed as: It smplfes for two ndependent varables: The sample ft parameter b 0, b 1, and b are used to

More information

Small Area Interval Estimation

Small Area Interval Estimation .. Small Area Interval Estmaton Partha Lahr Jont Program n Survey Methodology Unversty of Maryland, College Park (Based on jont work wth Masayo Yoshmor, Former JPSM Vstng PhD Student and Research Fellow

More information

Discussion of Extensions of the Gauss-Markov Theorem to the Case of Stochastic Regression Coefficients Ed Stanek

Discussion of Extensions of the Gauss-Markov Theorem to the Case of Stochastic Regression Coefficients Ed Stanek Dscusson of Extensons of the Gauss-arkov Theorem to the Case of Stochastc Regresson Coeffcents Ed Stanek Introducton Pfeffermann (984 dscusses extensons to the Gauss-arkov Theorem n settngs where regresson

More information

Small Area Estimation for Business Surveys

Small Area Estimation for Business Surveys ASA Secton on Survey Research Methods Small Area Estmaton for Busness Surveys Hukum Chandra Southampton Statstcal Scences Research Insttute, Unversty of Southampton Hghfeld, Southampton-SO17 1BJ, U.K.

More information

Chapter 5. Solution of System of Linear Equations. Module No. 6. Solution of Inconsistent and Ill Conditioned Systems

Chapter 5. Solution of System of Linear Equations. Module No. 6. Solution of Inconsistent and Ill Conditioned Systems Numercal Analyss by Dr. Anta Pal Assstant Professor Department of Mathematcs Natonal Insttute of Technology Durgapur Durgapur-713209 emal: anta.bue@gmal.com 1 . Chapter 5 Soluton of System of Lnear Equatons

More information

Chapter 8 Indicator Variables

Chapter 8 Indicator Variables Chapter 8 Indcator Varables In general, e explanatory varables n any regresson analyss are assumed to be quanttatve n nature. For example, e varables lke temperature, dstance, age etc. are quanttatve n

More information

STATISTICS QUESTIONS. Step by Step Solutions.

STATISTICS QUESTIONS. Step by Step Solutions. STATISTICS QUESTIONS Step by Step Solutons www.mathcracker.com 9//016 Problem 1: A researcher s nterested n the effects of famly sze on delnquency for a group of offenders and examnes famles wth one to

More information

Testing for seasonal unit roots in heterogeneous panels

Testing for seasonal unit roots in heterogeneous panels Testng for seasonal unt roots n heterogeneous panels Jesus Otero * Facultad de Economía Unversdad del Rosaro, Colomba Jeremy Smth Department of Economcs Unversty of arwck Monca Gulett Aston Busness School

More information

Lecture 9: Linear regression: centering, hypothesis testing, multiple covariates, and confounding

Lecture 9: Linear regression: centering, hypothesis testing, multiple covariates, and confounding Recall: man dea of lnear regresson Lecture 9: Lnear regresson: centerng, hypothess testng, multple covarates, and confoundng Sandy Eckel seckel@jhsph.edu 6 May 8 Lnear regresson can be used to study an

More information

Lecture 9: Linear regression: centering, hypothesis testing, multiple covariates, and confounding

Lecture 9: Linear regression: centering, hypothesis testing, multiple covariates, and confounding Lecture 9: Lnear regresson: centerng, hypothess testng, multple covarates, and confoundng Sandy Eckel seckel@jhsph.edu 6 May 008 Recall: man dea of lnear regresson Lnear regresson can be used to study

More information

Economics 130. Lecture 4 Simple Linear Regression Continued

Economics 130. Lecture 4 Simple Linear Regression Continued Economcs 130 Lecture 4 Contnued Readngs for Week 4 Text, Chapter and 3. We contnue wth addressng our second ssue + add n how we evaluate these relatonshps: Where do we get data to do ths analyss? How do

More information

Chapter 14 Simple Linear Regression

Chapter 14 Simple Linear Regression Chapter 4 Smple Lnear Regresson Chapter 4 - Smple Lnear Regresson Manageral decsons often are based on the relatonshp between two or more varables. Regresson analss can be used to develop an equaton showng

More information

SIMPLE LINEAR REGRESSION

SIMPLE LINEAR REGRESSION Smple Lnear Regresson and Correlaton Introducton Prevousl, our attenton has been focused on one varable whch we desgnated b x. Frequentl, t s desrable to learn somethng about the relatonshp between two

More information

Chapter 9: Statistical Inference and the Relationship between Two Variables

Chapter 9: Statistical Inference and the Relationship between Two Variables Chapter 9: Statstcal Inference and the Relatonshp between Two Varables Key Words The Regresson Model The Sample Regresson Equaton The Pearson Correlaton Coeffcent Learnng Outcomes After studyng ths chapter,

More information

Introduction to Regression

Introduction to Regression Introducton to Regresson Dr Tom Ilvento Department of Food and Resource Economcs Overvew The last part of the course wll focus on Regresson Analyss Ths s one of the more powerful statstcal technques Provdes

More information

Polynomial Regression Models

Polynomial Regression Models LINEAR REGRESSION ANALYSIS MODULE XII Lecture - 6 Polynomal Regresson Models Dr. Shalabh Department of Mathematcs and Statstcs Indan Insttute of Technology Kanpur Test of sgnfcance To test the sgnfcance

More information

Parametric fractional imputation for missing data analysis. Jae Kwang Kim Survey Working Group Seminar March 29, 2010

Parametric fractional imputation for missing data analysis. Jae Kwang Kim Survey Working Group Seminar March 29, 2010 Parametrc fractonal mputaton for mssng data analyss Jae Kwang Km Survey Workng Group Semnar March 29, 2010 1 Outlne Introducton Proposed method Fractonal mputaton Approxmaton Varance estmaton Multple mputaton

More information

The Ordinary Least Squares (OLS) Estimator

The Ordinary Least Squares (OLS) Estimator The Ordnary Least Squares (OLS) Estmator 1 Regresson Analyss Regresson Analyss: a statstcal technque for nvestgatng and modelng the relatonshp between varables. Applcatons: Engneerng, the physcal and chemcal

More information

Chapter 6. Supplemental Text Material

Chapter 6. Supplemental Text Material Chapter 6. Supplemental Text Materal S6-. actor Effect Estmates are Least Squares Estmates We have gven heurstc or ntutve explanatons of how the estmates of the factor effects are obtaned n the textboo.

More information

e i is a random error

e i is a random error Chapter - The Smple Lnear Regresson Model The lnear regresson equaton s: where + β + β e for,..., and are observable varables e s a random error How can an estmaton rule be constructed for the unknown

More information

ECONOMICS 351*-A Mid-Term Exam -- Fall Term 2000 Page 1 of 13 pages. QUEEN'S UNIVERSITY AT KINGSTON Department of Economics

ECONOMICS 351*-A Mid-Term Exam -- Fall Term 2000 Page 1 of 13 pages. QUEEN'S UNIVERSITY AT KINGSTON Department of Economics ECOOMICS 35*-A Md-Term Exam -- Fall Term 000 Page of 3 pages QUEE'S UIVERSITY AT KIGSTO Department of Economcs ECOOMICS 35* - Secton A Introductory Econometrcs Fall Term 000 MID-TERM EAM ASWERS MG Abbott

More information

UNR Joint Economics Working Paper Series Working Paper No Further Analysis of the Zipf Law: Does the Rank-Size Rule Really Exist?

UNR Joint Economics Working Paper Series Working Paper No Further Analysis of the Zipf Law: Does the Rank-Size Rule Really Exist? UNR Jont Economcs Workng Paper Seres Workng Paper No. 08-005 Further Analyss of the Zpf Law: Does the Rank-Sze Rule Really Exst? Fungsa Nota and Shunfeng Song Department of Economcs /030 Unversty of Nevada,

More information

RELIABILITY ASSESSMENT

RELIABILITY ASSESSMENT CHAPTER Rsk Analyss n Engneerng and Economcs RELIABILITY ASSESSMENT A. J. Clark School of Engneerng Department of Cvl and Envronmental Engneerng 4a CHAPMAN HALL/CRC Rsk Analyss for Engneerng Department

More information

Non-Mixture Cure Model for Interval Censored Data: Simulation Study ABSTRACT

Non-Mixture Cure Model for Interval Censored Data: Simulation Study ABSTRACT Malaysan Journal of Mathematcal Scences 8(S): 37-44 (2014) Specal Issue: Internatonal Conference on Mathematcal Scences and Statstcs 2013 (ICMSS2013) MALAYSIAN JOURNAL OF MATHEMATICAL SCIENCES Journal

More information

LINEAR REGRESSION ANALYSIS. MODULE IX Lecture Multicollinearity

LINEAR REGRESSION ANALYSIS. MODULE IX Lecture Multicollinearity LINEAR REGRESSION ANALYSIS MODULE IX Lecture - 30 Multcollnearty Dr. Shalabh Department of Mathematcs and Statstcs Indan Insttute of Technology Kanpur 2 Remedes for multcollnearty Varous technques have

More information

Learning Objectives for Chapter 11

Learning Objectives for Chapter 11 Chapter : Lnear Regresson and Correlaton Methods Hldebrand, Ott and Gray Basc Statstcal Ideas for Managers Second Edton Learnng Objectves for Chapter Usng the scatterplot n regresson analyss Usng the method

More information

Linear Approximation with Regularization and Moving Least Squares

Linear Approximation with Regularization and Moving Least Squares Lnear Approxmaton wth Regularzaton and Movng Least Squares Igor Grešovn May 007 Revson 4.6 (Revson : March 004). 5 4 3 0.5 3 3.5 4 Contents: Lnear Fttng...4. Weghted Least Squares n Functon Approxmaton...

More information

ANOVA. The Observations y ij

ANOVA. The Observations y ij ANOVA Stands for ANalyss Of VArance But t s a test of dfferences n means The dea: The Observatons y j Treatment group = 1 = 2 = k y 11 y 21 y k,1 y 12 y 22 y k,2 y 1, n1 y 2, n2 y k, nk means: m 1 m 2

More information

BIO Lab 2: TWO-LEVEL NORMAL MODELS with school children popularity data

BIO Lab 2: TWO-LEVEL NORMAL MODELS with school children popularity data Lab : TWO-LEVEL NORMAL MODELS wth school chldren popularty data Purpose: Introduce basc two-level models for normally dstrbuted responses usng STATA. In partcular, we dscuss Random ntercept models wthout

More information

Non-parametric bootstrap mean squared error estimation for M-quantile estimates of small area means, quantiles and poverty indicators *

Non-parametric bootstrap mean squared error estimation for M-quantile estimates of small area means, quantiles and poverty indicators * Non-parametrc bootstrap mean squared error maton for M-quantle mates of small area means quantles and poverty ndcators * Stefano Marchett 1 Monca Prates 2 Nos zavds 3 1 Unversty of Psa e-mal: stefano.marchett@for.unp.t

More information

Nonparametric model calibration estimation in survey sampling

Nonparametric model calibration estimation in survey sampling Ames February 18, 004 Nonparametrc model calbraton estmaton n survey samplng M. Govanna Ranall Department of Statstcs, Colorado State Unversty (Jont work wth G.E. Montanar, Dpartmento d Scenze Statstche,

More information

Bayesian predictive Configural Frequency Analysis

Bayesian predictive Configural Frequency Analysis Psychologcal Test and Assessment Modelng, Volume 54, 2012 (3), 285-292 Bayesan predctve Confgural Frequency Analyss Eduardo Gutérrez-Peña 1 Abstract Confgural Frequency Analyss s a method for cell-wse

More information

Numerical Heat and Mass Transfer

Numerical Heat and Mass Transfer Master degree n Mechancal Engneerng Numercal Heat and Mass Transfer 06-Fnte-Dfference Method (One-dmensonal, steady state heat conducton) Fausto Arpno f.arpno@uncas.t Introducton Why we use models and

More information

Efficient nonresponse weighting adjustment using estimated response probability

Efficient nonresponse weighting adjustment using estimated response probability Effcent nonresponse weghtng adjustment usng estmated response probablty Jae Kwang Km Department of Appled Statstcs, Yonse Unversty, Seoul, 120-749, KOREA Key Words: Regresson estmator, Propensty score,

More information

Feb 14: Spatial analysis of data fields

Feb 14: Spatial analysis of data fields Feb 4: Spatal analyss of data felds Mappng rregularly sampled data onto a regular grd Many analyss technques for geophyscal data requre the data be located at regular ntervals n space and/or tme. hs s

More information

DO NOT OPEN THE QUESTION PAPER UNTIL INSTRUCTED TO DO SO BY THE CHIEF INVIGILATOR. Introductory Econometrics 1 hour 30 minutes

DO NOT OPEN THE QUESTION PAPER UNTIL INSTRUCTED TO DO SO BY THE CHIEF INVIGILATOR. Introductory Econometrics 1 hour 30 minutes 25/6 Canddates Only January Examnatons 26 Student Number: Desk Number:...... DO NOT OPEN THE QUESTION PAPER UNTIL INSTRUCTED TO DO SO BY THE CHIEF INVIGILATOR Department Module Code Module Ttle Exam Duraton

More information

Composite Hypotheses testing

Composite Hypotheses testing Composte ypotheses testng In many hypothess testng problems there are many possble dstrbutons that can occur under each of the hypotheses. The output of the source s a set of parameters (ponts n a parameter

More information

A note on regression estimation with unknown population size

A note on regression estimation with unknown population size Statstcs Publcatons Statstcs 6-016 A note on regresson estmaton wth unknown populaton sze Mchael A. Hdroglou Statstcs Canada Jae Kwang Km Iowa State Unversty jkm@astate.edu Chrstan Olver Nambeu Statstcs

More information

Econ Statistical Properties of the OLS estimator. Sanjaya DeSilva

Econ Statistical Properties of the OLS estimator. Sanjaya DeSilva Econ 39 - Statstcal Propertes of the OLS estmator Sanjaya DeSlva September, 008 1 Overvew Recall that the true regresson model s Y = β 0 + β 1 X + u (1) Applyng the OLS method to a sample of data, we estmate

More information

ANOMALIES OF THE MAGNITUDE OF THE BIAS OF THE MAXIMUM LIKELIHOOD ESTIMATOR OF THE REGRESSION SLOPE

ANOMALIES OF THE MAGNITUDE OF THE BIAS OF THE MAXIMUM LIKELIHOOD ESTIMATOR OF THE REGRESSION SLOPE P a g e ANOMALIES OF THE MAGNITUDE OF THE BIAS OF THE MAXIMUM LIKELIHOOD ESTIMATOR OF THE REGRESSION SLOPE Darmud O Drscoll ¹, Donald E. Ramrez ² ¹ Head of Department of Mathematcs and Computer Studes

More information

Simulated Power of the Discrete Cramér-von Mises Goodness-of-Fit Tests

Simulated Power of the Discrete Cramér-von Mises Goodness-of-Fit Tests Smulated of the Cramér-von Mses Goodness-of-Ft Tests Steele, M., Chaselng, J. and 3 Hurst, C. School of Mathematcal and Physcal Scences, James Cook Unversty, Australan School of Envronmental Studes, Grffth

More information

LINEAR REGRESSION ANALYSIS. MODULE IX Lecture Multicollinearity

LINEAR REGRESSION ANALYSIS. MODULE IX Lecture Multicollinearity LINEAR REGRESSION ANALYSIS MODULE IX Lecture - 31 Multcollnearty Dr. Shalabh Department of Mathematcs and Statstcs Indan Insttute of Technology Kanpur 6. Rdge regresson The OLSE s the best lnear unbased

More information

3.1 Expectation of Functions of Several Random Variables. )' be a k-dimensional discrete or continuous random vector, with joint PMF p (, E X E X1 E X

3.1 Expectation of Functions of Several Random Variables. )' be a k-dimensional discrete or continuous random vector, with joint PMF p (, E X E X1 E X Statstcs 1: Probablty Theory II 37 3 EPECTATION OF SEVERAL RANDOM VARIABLES As n Probablty Theory I, the nterest n most stuatons les not on the actual dstrbuton of a random vector, but rather on a number

More information

Statistical Evaluation of WATFLOOD

Statistical Evaluation of WATFLOOD tatstcal Evaluaton of WATFLD By: Angela MacLean, Dept. of Cvl & Envronmental Engneerng, Unversty of Waterloo, n. ctober, 005 The statstcs program assocated wth WATFLD uses spl.csv fle that s produced wth

More information

Chat eld, C. and A.J.Collins, Introduction to multivariate analysis. Chapman & Hall, 1980

Chat eld, C. and A.J.Collins, Introduction to multivariate analysis. Chapman & Hall, 1980 MT07: Multvarate Statstcal Methods Mke Tso: emal mke.tso@manchester.ac.uk Webpage for notes: http://www.maths.manchester.ac.uk/~mkt/new_teachng.htm. Introducton to multvarate data. Books Chat eld, C. and

More information

Lecture Notes on Linear Regression

Lecture Notes on Linear Regression Lecture Notes on Lnear Regresson Feng L fl@sdueducn Shandong Unversty, Chna Lnear Regresson Problem In regresson problem, we am at predct a contnuous target value gven an nput feature vector We assume

More information

Exponential Type Product Estimator for Finite Population Mean with Information on Auxiliary Attribute

Exponential Type Product Estimator for Finite Population Mean with Information on Auxiliary Attribute Avalable at http://pvamu.edu/aam Appl. Appl. Math. ISSN: 193-9466 Vol. 10, Issue 1 (June 015), pp. 106-113 Applcatons and Appled Mathematcs: An Internatonal Journal (AAM) Exponental Tpe Product Estmator

More information

Basically, if you have a dummy dependent variable you will be estimating a probability.

Basically, if you have a dummy dependent variable you will be estimating a probability. ECON 497: Lecture Notes 13 Page 1 of 1 Metropoltan State Unversty ECON 497: Research and Forecastng Lecture Notes 13 Dummy Dependent Varable Technques Studenmund Chapter 13 Bascally, f you have a dummy

More information

Module 3 LOSSY IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur

Module 3 LOSSY IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur Module 3 LOSSY IMAGE COMPRESSION SYSTEMS Verson ECE IIT, Kharagpur Lesson 6 Theory of Quantzaton Verson ECE IIT, Kharagpur Instructonal Objectves At the end of ths lesson, the students should be able to:

More information

Properties of Least Squares

Properties of Least Squares Week 3 3.1 Smple Lnear Regresson Model 3. Propertes of Least Squares Estmators Y Y β 1 + β X + u weekly famly expendtures X weekly famly ncome For a gven level of x, the expected level of food expendtures

More information