An Adaptive Empirical Likelihood Test For Time Series Models 1

Size: px
Start display at page:

Download "An Adaptive Empirical Likelihood Test For Time Series Models 1"

Transcription

1 A Adaptive Empirical Likeliood Test For Time Series Models 1 By Sog Xi Ce ad Jiti Gao We exted te adaptive ad rate-optimal test of Horowitz ad Spokoiy (001 for specificatio of parametric regressio models to weakly depedet time series regressio models wit a empirical likeliood formulatio of our test statistic. It is foud tat te proposed adaptive empirical likeliood test preserves te rate-optimal property of te test of Horowitz ad Spokoiy (001. KEYWORDS: Empirical likeliood, goodess of fit test, kerel estimatio, rate-optimal test. 1. INTRODUCTION Cosider a time series eteroscedasticity regressio model of te form (1.1 Y t = m(x t + σ(x t e t, t = 1,,..., were bot m( ad σ( are ukow fuctios defied over R d, te data {(X t, Y t } t=1 are weakly depedet statioary time series, ad e t is a error process wit zero mea ad uit variace. Suppose tat {m θ ( θ Θ} is a family of parametric specificatio to te regressio fuctio m(x were θ R q is a ukow parameter belogig to a parameter space Θ. Tis paper cosiders testig te validity of te parametric specificatio of m θ (x agaist a series of local alteratives, tat is to test (1. H 0 : m(x = m θ (x versus H 1 : m(x = m θ (x + C (x for all x S, were C is a o-radom sequece tedig to zero as, (x is a sequece of fuctios i R d ad S is a compact set i R d. Bot C ad (x caracterize te departure of te local alterative family of regressio models from te parametric family {m θ ( θ Θ}. Tere ave bee extesive ivestigatios o employig te kerel smootig metod to form oparametric specificatio tests for a ull ypotesis like H 0 ; see Härdle ad Mamme (1993, Hjellvik, Yao ad Tjøsteim (1998 ad oters. A commo feature amog tese tests is tat te test statistics are formulated based o a sigle kerel smootig badwidt wic coverges to 0 as. Tis leads to a commo cosequece tat C, wic defies te gap betwee H 0 ad H 1, as to be at least of te order of 1/ d/4 i order to ave cosistet tests. I oter words, tese tests are uable to distiguis betwee H 0 ad H 1 for C at a order smaller ta 1/ d/4, wic ca be muc larger ta 1/, te order acieved by some oter oparametric tests for te above H 0 versus H 1 wit (x (x, for istace te coditioal Kolmogorov test cosidered i Adrews (1997. Te sigle badwidt based kerel tests also as o built-i adaptability to te smootess of (. 1 We would like to tak Mr Ceg Yog Tag for valuable computatio assistace wo was supported uder a Natioal Uiversity of Sigapore researc grat R Te secod autor ackowledges support from a Australia Researc Coucil Discovery Grat. 1

2 I a sigificat developmet, Horowitz ad Spokoiy (001 propose a test by combiig a studetized versio of te kerel based test statistic of Härdle ad Mamme (1993 over a set of badwidts. Tey establis i te more restrictive case of (x (x tat te test is cosistet for C = O( 1/ log log( wic almost reaces te order of 1/. Most remarkably, te test is adaptive to te ukow smootess of m i H 1 wic forms a Hölder smootess class of fuctios. I particular, if fuctios i te Hölder class possess a ukow s-order derivatives, te test is cosistet for C = O{( 1 log log( s/(4s+d } for s max (, d 4, wic is te optimal rate of covergece for C i te miimax sese of Spokoiy (1996 ad Igster ad Suslia (003. We cosider i tis paper two extesios of te adaptive test of Horowitz ad Spokoiy. Oe is to iclude weakly depedet observatios; te oter is to use te empirical likeliood (EL of Owe (1988 to formulate te test statistic, wic is desiged to equip te test statistic wit some favorable features of te EL. We sow tat te above metioed optimal or ear optimal rates for C establised by Horowitz ad Spokoiy (001 are maitaied uder tese extesios. Te rest of tis paper is orgaized as follows. Sectio proposes te adaptive empirical likeliood test ad presets te rate-optimal property of te test. Sectio 3 presets simulatio results. All te tecical proofs are provided i te appedix.. ADAPTIVE EMPIRICAL LIKELIHOOD TEST Like existig kerel based goodess-of-fit tests, our test is based o a kerel estimator of te coditioal mea fuctio m(x. Let K be a d-dimesioal bouded fuctio wit a compact support o [ 1, 1] d. Let be a smootig badwidt satisfyig (.1 ad K (u = d K(u/. 0 ad d / log 6 ( as, Te Nadaraya-Watso (NW estimators of m(x is ˆm(x = t=1 K (x X t Y t t=1 K (x X i. Let θ be a cosistet estimator of θ uder H 0. Like Härdle ad Mamme (1993, let m θ(x t=1 = K (x X t m θ(x t t=1 K (x X t be a kerel smoot of te parametric model m θ (x wit te same kerel ad badwidt as i ˆm(x. Tis is to avoid te bias of te kerel estimator i te goodess-of-fit test. Ce, Härdle ad Li (003 propose a test statistic based o te EL as follows. Let Q t (x = K (x X t {Y t m θ(x}. At a arbitrary x S, let p t (x be oegative real Oter kerel based EL tests wit a sigle badwidt are Fa ad Zag (004 ad Tripati ad Kitamura (004.

3 umbers represetig weigts allocated to eac (X t, Y t. Te EL for m(x evaluated at te smooted parametric model m θ(x is (. L{ m θ(x} = max p t (x t=1 subject to t=1 p t(x = 1 ad t=1 p t(xq t (x = 0. As te EL is maximized at p t (x = 1, te log-el ratio is l{ m θ(x} = log[l{ m θ(x} ]. Te EL test statistic at a give badwidt is (.3 l( m θ; = l{ m θ(x}π(xdx, were π( is a o-egative weigt fuctio supported o te compact set S R d satisfyig π(xdx = 1 ad π (xdx <. Let R(K = K (xdx, v(x = R(Kσ (xf 1 (x ad (K C(K, π = R (K ( (x (.4 dx π (ydy, were K ( is te covolutio of K. Ce, Härdle ad Li (003 sow tat as { } d/ l( m θ; 1 d/ v 1/ (x (xπ(xdx d (.5 N(0, C(K, π for te case of π(x = S 1 I(x S were I is te idicator fuctio ad S is te volume of S. A extesio of (.5 to a geeral weigt fuctio is automatic. Ce, Härdle ad Li (003 te proposes a sigle badwidt based EL test based o critical values obtaied by simulatig a Gaussia radom field. Like all oparametric kerel goodess-of-tests based o a sigle badwidt, te test is cosistet oly if C is at te order of 1/ d/4 or larger, idicatig tat C as to coverge to zero more slowly ta 1/. Te latter is te rate establised for oparametric goodessof-fit tests based o te residuals we tere is o smootig ivolved. To reduce te order of C, we employ te adaptive test procedure of Horowitz ad Spokoiy (001 for te EL test as follows. Let H = { } (.6 = max a k : mi, k = 0, 1,,... J be a set of badwidts, were 0 < a < 1, J = log 1/a ( max / mi is te umber of badwidts i H, max = c max (log log( 1 d ad mi = c mi γ for 0 < γ < 1 ad some positive d costats < c mi, c max <. Te coice of max is vital i reducig C to almost 1/ rate i te case of ( (. Te rage of γ allows mi = O{ 1/(4+d }, te optimal 3

4 order i te kerel estimatio of m(x. I view of te fact tat E{l( m θ; } = 1 uder H 0 ad var{l( m θ; } = C(K, π d as give i (.4 te adaptive EL test statistic is proposed as follows: (.7 l( L = max m θ; 1. H C(K, π d Let l α (0 < α < 1 be te 1 α quatile of L were α is te sigificace level of te test. Motivated by te bootstrap procedure of Horowitz ad Spokoiy, we propose te followig bootstrap procedure to approximate l α : 1. For eac t = 1,,...,, let Yt = m θ (X t + σ (X t e t, were σ ( is a cosistet estimator of σ( ad {e t} is sampled radomly from a distributio wit E[e t] = 0, E [e t ] = 1 ad E [ e t 4+δ] < for some δ > 0.. Let ˆθ be te estimate of θ based o te resample {(X t, Yt } t=1. Compute te statistic L by replacig Y t ad θ wit Yt ad ˆθ accordig to ( Estimate l α by l α, te 1 α quatile of te empirical distributio of L, wic ca be obtaied by repeatig steps 1 may times. Te estimator σ ( ca be te followig kerel estimator (.8 σ (x = t=1 K b(x X t {Y t ˆm(x} t=1 K b(x X t wit a badwidt b suc tat mi b d as. Te proposed adaptive EL test rejects H 0 if L > l α. 3. MAIN RESULTS Te followig teorem sows tat te adaptive EL test as a correct size asymptotically. Teorem 3.1. Suppose Assumptios A.1 ad A.(i(ii(iv old. Te uder H 0, lim P (L > lα = α. I te followig, we establis te cosistecy of te adaptive EL test agaist a sequece of fixed, local ad smoot alteratives, respectively. Let te parameter space Θ be a ope subset of R q. Let M = {m θ ( : θ Θ} ad f(x be te margial desity of X i. We ow defie te distace betwee m ad te parametric family M as [ ( 1/ (3.1 ρ(m, M = if [m θ (x m(x] f(xdx]. θ Θ x S Te cosistecy of te test agaist a fixed alterative is establised i Teorem 3. below. 4

5 Teorem 3.. Assume tat Assumptios A.1 ad A.(i(iii(iv old. If tere is a C ρ > 0 suc tat ρ(m, M C ρ for 0 wit some large 0, te lim P (L > l α = 1. We te cosider te cosistecy of te EL test agaist special from of H 1 of te form (3. m(x = m θ (x + C (x were C 0 as, θ Θ ad for positive ad fiite costats D 1, D ad D 3, (3.3 0 < D 1 (xf(xdx D < ad ρ(m, M D 3 C. x S Teorem 3.3. Assume Assumptios A.1 ad A.(i(iii. Let Assumptio A.(iv old wit max = C (log log( 1 d for some fiite costat C. Let m satisfy (3. ad (3.3 wit C C 1/ log log( for some costat C > 0. Te lim P (L > lα = 1. To discuss te cosistecy of te adaptive EL test over alteratives i a Hölder smootess class, we itroduce te followig otatio. Let j = (j 1,..., j d were j 1,..., j d 0 are itegers, j = d k=1 j d ad D j m(x = j m(x weever te derivative exists. Defie te x j 1 1 x j d d Hölder orm m H,s = sup x S j s ( Dj m(x. Te smootess class tat we cosider cosist of fuctios m S(H, s {m : m H,s C H } for some ukow s ad C H <. For s max(, d/4 ad all sufficietly large D m <, defie (3.4 B H, = ( {m S(H, s : ρ(m, M D m 1 } s/(4s+d log log(. Teorem 3.4. Assume tat Assumptios A.1 ad A. old. Let m satisfy (1. uder H 1 ad (3.4. Te for 0 < α < 1 ad B H, defied i (3.4, lim if m BH, P (L > l α = SIMULATION RESULTS We carried out two simulatio studies wic were desiged to evaluate te empirical performace of te proposed adaptive EL test. I te first simulatio study, we coducted simulatio for te followig regressio model used i Horowitz ad Spokoiy (001: (4.1 Y i = β 0 + β 1 X i + (5/τφ(X i /τ + ɛ i, were {ɛ i } are idepedet ad idetically distributed from tree distributios wit zero mea ad costat variace, {X i } are uivariate desig poits ad θ = (β 0, β 1 τ = (1, 1 τ is cose as te true vector of parameters ad φ is te stadard ormal desity fuctio. Te ull ypotesis H 0 : m(x = β 0 + β 1 x specifies a liear regressio correspodig to τ = 0, wereas te alterative ypotesis H 1 : m(x = β 0 + β 1 x + (5/τφ(x/τ for τ = 1.0 5

6 ad 0.5. Readers sould refer to Horowitz ad Spokoiy (001 for details o te desigs X i, te tree distributios of ɛ i ad oter aspects of te simulatio. We used te same umber of simulatio, te bootstrap resamples ad estimatio procedures for θ as i Horowitz ad Spokoiy (001. We also employed te same kerel, te same badwidt set H, ad te same estimator σ ad te distributio for e i i te bootstrap procedure as i Horowitz ad Spokoiy (001. Like Horowitz ad Spokoiy, te omial size of te test was 5%. Table 1 summaries te performace of te adaptive EL test by addig oe colum to Table 1 of Horowitz ad Spokoiy (001. Our results sow tat te proposed adaptive EL test as sligtly better power ta te adaptive test of Horowitz ad Spokoiy (001, wile te sizes are similar to tose of Horowitz ad Spokoiy (001. Tis may ot be surprisig as te two tests are equivalet i te first order. Te differeces betwee te two tests are (i te EL test statistic carries out te studetizig implicitly ad (ii certai iger order features like te skewess ad kurtosis are reflected i te EL statistic. Tese migt be te uderlyig cause for te sligtly better power observed for te EL test. Te secod simulatio study was coducted o a ARCH type time series regressio model of te form: (4. Y i = Y i 1 + C cos(8y i Yi e i, were te iovatio {e i } i=1 was cose to be idepedet ad idetically distributed N(0, 1 radom variables. Te sample sizes cosidered i te simulatio were = 300 ad = 500. Te vector of parameters θ = (α, β, σ was estimated usig te pseudo-maximum likeliood metod, wic is commoly used i te estimatio of ARCH models. I te bootstrap implemetatio, we cose e iid i N(0, 1 ad te estimator σ(x give i (.8. We cose te badwidt set H = {0.3, 0.33, 0.367, 0.407, 0.45} wit a = for = 300 ad H = {0.5, 0.81, 0.316, 0.356, 0.4} wit a = for = 500. Bot te power ad te size of te adaptive test are reported i Table. We foud te test ad good approximatio to te omial sigficiace level of 5%, wic cofirms Teorem 3.1 ad te quality of te bootstrap calibratio to te distributio of te adaptive EL test statistic. As expected we C was icreased, te power of te test was icreased; ad for a fixed level of C, te power icreased we was icreased. Te latter was because te distace betwee H 0 ad H 1 became larger we was icreased altoug C was kept te same. Departmet of Statistics, Iowa State Uiversity, Ames, IA 50010; sogce@iastate.edu; ttp:// ad Scool of Matematics ad Statistics, Te Uiversity of Wester Australia, Crawley WA 6009, Australia; jiti@mats.uwa.edu.au; ttp:// jiti 6

7 APPENDIX Tis appedix gives te assumptios ad proofs of te teorems give i Sectio 3. A.1. Assumptios Assumptio A.1. (i Assume tat te process (X t, Y t is strictly statioary ad α-mixig wit te mixig coefficiet α(t = sup{ P (A B P (AP (B : A Ω s 1, B Ω s+t } for all s, t 1, were Ω j i deotes te σ-field geerated by {(X s, Y s : i s j}. Tere exist costats a > 0 ad ρ [0, 1 suc tat α(t aρ t for t 1. (ii Assume tat for all t 1, P (E[e t Ω t 1 ] = 0 = 1 ad P ( E[e t Ω t 1] = 1 = 1, were Ω t = σ{(x s+1, Y s : 1 s t} is a sequece of σ-fields geerated by {(X s+1, Y s : 1 s t}. (iii Let ɛ t = Y t m(x t. Tere exists a positive costat δ > 0 suc tat E [ ɛ t 4+δ] <. (iv Let µ i (x = E[ɛ i t X = x], S π be a compact subset of R d, S f be te support of f ad S = S π S f be a compact set i R d. Let π be a weigt fuctio supported o S suc tat tat s S π(sds = 1 ad s S π (sds C for some costat C. I additio, te margial desity fuctio, f(x, of X t ad µ i (x for i = or 4 are all Lipscitz cotiuous i S, ad tat all te first two derivatives of f(x, m(x ad µ (x are cotiuous o R d, if x S σ(x C 0 > 0 for some costat C 0, ad o S te desity fuctio f(x is bouded below by C f ad above by C 1 f for some C f > 0. (v Te kerel K is a product kerel as defied by K(x 1,, x d = d i=1 k(x i, were k( is a r-t order uivariate kerel wic is symmetric ad supported o [ 1, 1], ad satisfies k(tdt = 1, t l k(tdt = 0 for l = 1,, r 1 ad t r k(tdt = k r 0 for a positive iteger r > d/. I additio, k(x is Lipscitz cotiuous i [ 1, 1]. Let te parameter set Θ be a ope subset of R q. Let M = {m θ ( : θ Θ}. Defie θ m θ (x = m θ (x θ, θ m θ(x = m θ (x θ θ, ad 3 θ m θ(x = 3 m θ (x θ θ θ weever tese derivatives exist. For ay q q matrix D, defie D = sup Dv v R q v, were v = q i=1 v i for v = (v 1,..., v q τ. Assumptio A.. (i Te parameter set Θ is a ope subset of R q for some q 1. Te parametric family M = {m θ ( : θ Θ} satisfies: For eac x S, m θ (x is twice differetiable almost surely wit respect to θ Θ. Assume tat tere are costats 0 < C 1, C < suc tat [ [ ] E ] sup m θ (X 1 C 1 ad max E θ Θ 1 j 3 were B m = q i=1 q j=1 b ij for B = {b ij} 1 i,j q. sup j θ m θ(x 1 m θ Θ C, For eac θ Θ, m θ (x is cotiuous wit respect to x R d. Assume tat tere is a fiite C I > 0 suc tat for every ε > 0 if [m θ(x m θ (x] f(xdx C I ε. θ,θ Θ: θ θ ε x S 7

8 ( (ii Let H 0 be true. Te θ 0 Θ ad lim P θ θ0 > C L < ε for ay ε > 0 ad all sufficietly large C L. ( (iii Let H 0 be false. Te tere is a θ Θ suc tat lim P θ θ > C L < ε for ay ε > 0 ad all sufficietly large C L. (iv Assume tat te set H as te structure of (.6 wit max > mi γ for some costat γ suc tat 0 < γ < 1 d ad max = C (log log( 1 d for some fiite costat C > 0. Assumptio A.1 is quite stadard i tis kid of problem ad Assumptio A. correspods to Assumptios 1, 4 ad 6 of Horowitz ad Spokoiy (001. A.. Tecical Lemmas From (.6 of Ce, Härdle ad Li (003, oe may sow tat (A.1 l( m θ ; = d Ū1 (x; θv 1 (xπ(xdx + o p ( d/ uiformly i H, were v(x = R(Kf 1 (xσ (x if te issue of boudary is ot cosidered, ad Ū 1 (x; θ ( x = ( d 1 Xt K {Y t m θ (x}. t=1 Let W t (x = 1 K ( x X t d, ast = d x S W s(xw t (xv 1 (xπ(xdx, ad λ t (θ = λ(x t, θ = m(x t m θ (X t. Defie (A. l 0 ( = s,t a st ɛ s ɛ t ad Q (θ = Q (θ; = s,t a st λ s (θλ t (θ. Te te leadig term i l ( m θ; is (A.3 l 1 (, θ d Ū1 (x; θv 1 (xπ(xdx = l 0 ( + Q ( θ + Π ( θ, were Π ( θ = l 1 (; θ l 0 ( Q ( θ is te remaider term. Witout loss of geerality, we assume tat C(K, π = R (K ( K ( (x dx π (ydy = 1. I view of te defiitio of L = max H l( m θ; 1 d/ (A.4 ad (A.3, defie L 0 ( = l 0( 1 d/, L 1 ( = l 1(, θ 1 d/ ad L ( = l 1(, θ 1 d/, were θ = θ 0 we H 0 is true ad θ is as defied i Assumptio A.(iii we H 0 is false. Let L 0 ( ad L 1 ( be te correspodig versios of L 0( ad L 1 ( wit {(X i, Y i } ad θ replaced by {(X i, Yi } ad ˆθ respectively. Te followig two Lemmas are preseted witout proof ere, wic ca be foud from te proofs of Lemmas A.1 ad A.6 of Ce ad Gao (004. 8

9 Lemma A.1. Suppose tat Assumptios A.1 ad A. old. For eac θ Θ ad sufficietly large, we ave tat C 1 d λ(θ τ λ(θ Q (θ C d λ(θ τ λ(θ olds i probability, were λ(θ = (λ 1 (θ,, λ (θ τ ad 0 < C 1 C < are costats. Lemma A.. Suppose tat Assumptios A.1 ad A. old. Te as max L ( H = max L 1 ( + o p (1 = max L ( + o p (1, H H max L 1 ( H = max L 0 ( + o p(1, H ad max H L 1 ( = max H L 0 ( + o p (1 uder H 0. Lemma A.3. Suppose Assumptios A.1 ad A. old. Te te asymptotic distributios of max H L ( ad max H L 0 ( are idetical uder H 0. Proof: I view of Lemma A., i order to prove Lemma A.3, it suffices to sow tat te distributios of max H L 0 ( ad max H L 0 ( are asymptotically te same. Similarly to te proof of Lemma A., we ca sow tat ( ( (A.5 max d/ a ss ɛ s 1 = o p (1 ad max d/ a ss ɛ s 1 = o p (1. H H s=1 Tus, it suffices to sow tat max H s t a stɛ s ɛ t ad max H s t a stɛ sɛ t ave te same asymptotic distributio. For H, let u t = ɛ t or ɛ t ad defie (A.6 s=1 B (u 1,..., u = d/ a st u s u t s t Let B (u 1,..., u be te sequece obtaied by stackig te correspodig B (u 1,..., u ( H. Let G( = G ( be a 3 times cotiuously differetiable fuctio over R J. Defie C (G = sup max 3 G(v v R J i,j,k=1,,...,j v i v j v k. Like Horowitz ad Spokoiy (001, tere are two steps i te proof of Lemma A.3. First, we wat to sow tat ( J E [G(B (ɛ 1,..., ɛ ] E [G(B (ɛ 1,..., ɛ 3 1/ (A.7 ] C 0 C (G for ay 3 times differetiable G(, some fiite costat C 0, ad all sufficietly large. Te i te secod step, (A.7 is used to sow tat B (ɛ 1,..., ɛ ad B (ɛ 1,..., ɛ ave te same asymptotic distributio. Trougout te rest of te proof, we replace a st i (A.6 wit ã st ( = d/ a st. Note tat (A.8 E [G(B (ɛ 1,..., ɛ ] E [G(B (ɛ 1,..., ɛ ] 9

10 E [ G(B (ɛ 1,..., ɛ t, ɛ t+1,, ɛ ] E [G(B (ɛ 1,..., ɛ t 1, ɛ t,..., ɛ ], t=1 were B (ɛ 1,..., ɛ, ɛ +1 = B (ɛ 1,..., ɛ ad B (ɛ 0, ɛ 1,..., ɛ = B (ɛ 1,..., ɛ. We ow derive a upper boud o te last term of te sum o te rigt ad side of (A.8. Similar bouds ca be derived for te oter terms. Let U 1, Λ ad Λ, respectively, deote te vectors tat are obtaied by stackig U, = 1 1 s=1 t=1, s ã st (ɛ s ɛ t, 1 Λ, = ɛ ã s (ɛ s, s=1 1 Λ, = ɛ ã s (ɛ s. s=1 Usig a Taylor expasio to te last term of te sum o te rigt ad side of (A.8 about ɛ = ɛ = 0 gives [ E [G(B (ɛ 1,..., ɛ ] E [G(B (ɛ 1,..., ɛ 1, ɛ ] E G (U 1 (Λ Λ ] + 1 [ ] E Λ τ G (U 1 Λ Λ τ G C (G { (U 1 Λ + E [ Λ 3] [ + E Λ 3]}, 6 were G ad G deote te gradiet ad matrix of secod derivatives of G ad C (G is a positive ad fiite costat. ] [ ] Sice E [ɛ j Ω 1 = E ɛ j Ω 1 for j = 1,, we ave Tis implies (A.9 E [(Λ Λ ] Ω 1 = 0 ad E [(Λ Λ τ Λ ] Λτ Ω 1 = 0. E [G(B (ɛ 1,..., ɛ ] E [G(B (ɛ 1,..., ɛ 1, ɛ ] C (G 6 { E [ Λ 3] [ + E Λ 3]}. To estimate te upper boud of (A.9, we eed te followig result: ( ( 1 x Xs x Xt (A.10 a st = d K K q(xdx = 1 ( K(uK u + X s X t q(x s + udu = 1 ( L Xs X t, X s, were q(x = v 1 (xπ(x ad L (x, y = K(uK(u + xq(y + udu. Usig Assumptios A.1 A. ad (A.10, we ave as (A ã s ( 1ã t ( ɛ s ɛ t ɛ4 E 1 H H s=1 t=1, s 10

11 = = 1 H 1 H [ ( E L Xs X ( ] 1 d, X s L Xt X H 1 d, X t ɛ 1 sɛ tɛ 4 ( ( ( 1 d x z y z L, x L 1 H 1 L (x 1, x 3 + x 1 1 L (x, x 3 + x u v w 4 H 1 H, y u v w 4 f(x, y, z, u, v, wdxdydzdudvdw f(x 3 + x 1 1, x 3 + x, x 3, u, v, wdx 1 dx dx 3 dudvdw C ( J (1 + o(1, were f(x, y, z, u, v, w is te joit desity fuctio of (X s, X t, X, ɛ s, ɛ t, ɛ ad 0 < C < is a costat. Similarly to te proof of Lemma C. of Gao ad Kig (001, we ca sow tat as 1 ( 1 d E ( a s( 1 a s ( a t ( ɛ 3 sɛ t ɛ 4 J (A.1 = o, 1, H 1 s t 1 1 ( 1 d E ( a s( 1 a t ( a u ( ɛ sɛ t ɛ u ɛ 4 J = o, 1, H 1 s t,s u,t u q 1 ( 1 d E ( a s ( 1 a t ( 1 a u ( a v ( ɛ s ɛ t ɛ u ɛ v ɛ 4 J = o 1, H 1 1 s t,s u,s v,t u,t v,u v 1 usig Assumptio A.1(ii tat for every give x, [ ( ] [ ( ] Xt x Xt x (A.13 E L, X t ɛ t = E L, X t E [ɛ t Ω t 1 ] Equatios (A.11 ad (A.1 te imply tat as 1 ( ã s ( 1 ɛ s ã t ( 1 ɛ t ã u ( ɛ u ã v ( ɛ v ɛ 4 J (A.14 C. E 1 H H s,t,u,v=1 Let à s be te vector tat is obtaied by stackig ã s ( ( H. Equatio (A.14 te implies tat as (A.15 E [ Λ 3] 1 = 8E 3 à s ɛ s ɛ 8 E = 8 { s=1 E 1 H H [ 1 s,t,u,v=1 H ( 1 = 0. ã s (ɛ s ɛ s=1 3/4 ã s ( 1 ɛ s ã t ( 1 ɛ t ã u ( ɛ u ã v ( ɛ v ɛ 4 ]} 3/4 C ( J 3/. 11

12 ] A similar result olds for E [ Λ 3. Tus (A.16 E [ Λ 3] [ + E Λ 3] C ( 3/ J. Step : Followig te lies of Horowitz ad Spokoiy (001 by utilizig te above establised boud (A.16 ad usig (A.8, it ca be sow tat as (A.17 ] [max P B (ɛ 1,..., ɛ x P H [ ] max B (ɛ 1,..., ɛ x C H Tis implies (A.7 ad fially completes te proof of Lemma A.3. ( J 3 1/ 0. Lemma A.4. Suppose tat Assumptio A.1 olds. Te for ay x 0, H ad all sufficietly large P (L 0( > x exp ( x. 4 Proof: Te proof is give as tat of Lemma A.8 i Ce ad Gao (004. For 0 < α < 1, defie l α to be te 1 α quatile of max H L 0 (. Lemma A.5. Suppose tat Assumptio A.1 olds. Te for large eoug l α log(j log(α. Proof: Te proof is similar to tat of Lemma 1 of Horowitz ad Spokoiy (001. Lemma A.6. Suppose tat Assumptios A.1 ad A. old. Suppose tat ( (A.18 lim P Q (θ d/ l α = 1 ( for some H, were l α = max l α, log(j + log(j. Te lim P (L > lα = 1. Proof: By (A., (A.3, (A.4 ad Lemma A., L ca be replaced wit max H L (. By Lemmas A. ad A.3, l α ca be replaced by l α. Tus, it suffices to sow tat lim P (max L ( > l α = 1, H wic olds if lim P (L ( > l α = 1 for some H. For ay H, usig (A., (A.3, (A.4 ad Lemma A. agai we ave (A.19 L ( = L 0 ( + d/ Q (θ + d/ Π (θ = L 0( + d/ Q (θ + d/ Π (θ + o p (1 = L 0( + d/ Q (θ (1 + o p (1 + o p (1. 1

13 Coditio (A.18 implies tat as (A.0 P (Q (θ < d/ l α 0. Observe tat P (L ( > l α = P (L ( > l α, Q (θ d/ l α ( + P L ( > l α, Q (θ < d/ l α I 1 + I. Tus, it follows from (A.19 tat as I 1 = P (L 0( + d/ Q (θ + d/ Π (θ > l α Q (θ d/ l α P (Q (θ d/ l α = P (L 0( + d/ Q (θ (1 + o p (1 > l α Q (θ d/ l α P (Q (θ d/ l α P (L 0( > l α l α Q (θ d/ l α P (Q (θ d/ l α 1 because L 0 ( is asymptotically ormal ad terefore bouded i probability ad l α l α as. Because of (A.0, lim I P (Q (θ < d/ l α = 0. Tis fiises te proof. A.3. Proofs of Teorems Proof of Teorem 3.1: By Lemma A., max H L 1 ( = max H L ( + o p (1. By Lemma A.3, uder H 0, max H L ( max H L 0 ( 0 i distributio as. Usig Lemma A. agai implies max H L 1 ( = max H L 0 ( + o p(1. Tis implies tat max H L 1 ( max H L 1 ( 0 i distributio as. Tis, alog wit equatios (A.1 (A.4, fiises te proof. I order to prove Teorems , i view of Lemma A.6, it suffices to verify (A.18. Usig Lemma A.1, it suffices to verify ( (A.1 lim P d λ(θ τ λ(θ 4 l α d/ = 1. Proof of Teorem 3.: I view of te defiitio of l α, equatio (A.1 follows from te fact tat as, 1 (A. λ(θτ λ(θ ρ(m, M 0 olds i probability ad d C 0 l α d/ for some costat 0 < C 0 < ad large eoug. (A.3 Proof of Teorem 3.3: Usig te defiitio of l α, (A., 1 [ (X t E S (X 1 ] = t=1 x S 13 (xf(xdx D 1 > 0 as,

14 ad te fact tat (A.4 1 λ(θτ λ(θ = C (X t D 1 C t=1 olds i probability, oe ca see tat (A.1 olds we = max = (log log( 1 d. Tis fiises te proof of Teorem 3.3. Proof of Teorem 3.4: I order to verify (A.18, we eed to itroduce te followig otatio: ( 1 = 1 l α 4s+d. Tis implies 4s+d 1 = l α. Coose H suc tat 1 < 1. We te ave 4 d l α = 4 d 4s+d 1 4 4s+d + (A.5 d = 4 s+d. (A.6 Tus, i order to verify (A.18, it suffices to sow tat Q (θ 4 s+d olds i probability for te selected H ad θ Θ. Te verificatio of (A.6 ca be doe usig similar teciques detailed i te proof of Lemma A.1 give i Ce ad Gao (004. Alteratively, oe may follow te proof of (A13 of Horowitz ad Spokoiy (001 by otig tat all te derivatios below teir (A13 old i probability for radom X i. REFERENCES Adrews, D. W. K. (1997: A Coditioal Kolmogorov Test, Ecoometrica, 65, Ce, S. X., ad J. Gao (004: A Adaptive Empirical Likeliood Test for Time Series Regressio Models, mauscript. Ce, S. X., W. Härdle, ad M. Li (003: A Empirical Likeliood Goodess of Fit Test for Time Series, Joural of te Royal Statistical Society, Series B., 65, Fa, J., ad J. Zag (004: Sieved Empiricial Likeliood Ratio Tests for Noparametric Fuctios, Te Aals of Statistics, 9, Gao, J. ad M. L. Kig (001: Estimatio ad Model Specificatio Testig i Noparametric ad Semiparametric Regressio, Workig paper available at jiti/jems.pdf. Härdle, W., ad E. Mamme (1993: Comparig Noparametric versus Parametric Regressio Fits, Te Aals of Statistics, 1, Hjellvik, V., Q. Yao, ad D. Tjøsteim (1998: Liearity Testig Usig Local Polyomial Approximatio, Joural of Statistical Plaig ad Iferece, 68, Horowitz, J. L., ad V. G. Spokoiy (001: A Adaptive, Rate Optimal Test of a Parametric Mea Regressio Model Agaist a Noparametric Alterative, Ecoometrica, 69, Igster, Yu. I. ad Suslia, I. A. (003: Noparametric Goodess-of-Fit Testig Uder Guassia Models, Lecture Notes i Statistics, 169, New York: Spriger-Verlag. Owe, A. (1988: Empirical Likeliood Ratio Cofidece Itervals for a Sigle Fuctioal, Biometrika, 75, Spokoiy, V. G. (1996: Adaptive Hypotesis Testig Usig Wavelets, Te Aals of Statistics, 4, Tripati, G., ad Y. Kitamura (004: O Testig Coditioal Momet Restrictios: te Caoical Case, Te Aals of Statistics (to appear. 14

15 TABLE 1 SIMULATION RESULTS ON MODEL (4.1 Probability of Rejectig Null Hypotesis Adrews Härdle-Mamme Horowitz-Spokoiy EL Distributio ɛ τ Test Test Test Test Null Hypotesis Is True Normal Mixture Extreme Value Null Hypotesis Is False Normal Mixture Extreme Value Normal Mixture Extreme Value TABLE SIMULATION RESULTS ON MODEL (4. C = 300 = 500 Null Hypotesis Alterative Hypotesis Alterative Hypotesis

LECTURE 2 LEAST SQUARES CROSS-VALIDATION FOR KERNEL DENSITY ESTIMATION

LECTURE 2 LEAST SQUARES CROSS-VALIDATION FOR KERNEL DENSITY ESTIMATION Jauary 3 07 LECTURE LEAST SQUARES CROSS-VALIDATION FOR ERNEL DENSITY ESTIMATION Noparametric kerel estimatio is extremely sesitive to te coice of badwidt as larger values of result i averagig over more

More information

Nonparametric regression: minimax upper and lower bounds

Nonparametric regression: minimax upper and lower bounds Capter 4 Noparametric regressio: miimax upper ad lower bouds 4. Itroductio We cosider oe of te two te most classical o-parametric problems i tis example: estimatig a regressio fuctio o a subset of te real

More information

Semiparametric Mixtures of Nonparametric Regressions

Semiparametric Mixtures of Nonparametric Regressions Semiparametric Mixtures of Noparametric Regressios Sijia Xiag, Weixi Yao Proofs I tis sectio, te coditios required by Teorems, 2, 3 ad 4 are listed. Tey are ot te weakest sufficiet coditios, but could

More information

Department of Econometrics and Business Statistics

Department of Econometrics and Business Statistics ISSN 440-77X Australia Departmet of Ecoometrics ad Busiess Statistics ttp://www.buseco.moas.edu.au/depts/ebs/pubs/wpapers/ A Improved Noparametric Uit Root Test Jiti Gao ad Maxwell ig August 0 Workig Paper

More information

4 Conditional Distribution Estimation

4 Conditional Distribution Estimation 4 Coditioal Distributio Estimatio 4. Estimators Te coditioal distributio (CDF) of y i give X i = x is F (y j x) = P (y i y j X i = x) = E ( (y i y) j X i = x) : Tis is te coditioal mea of te radom variable

More information

Bandwidth Selection in Nonparametric Kernel Testing

Bandwidth Selection in Nonparametric Kernel Testing Bwidt Selectio i Noparametric Kerel Testig Jiti GAO Irèe GIJBELS We propose a soud approac to bwidt selectio i oparametric kerel testig. Te mai idea is to fid a Edgewort expasio of te asymptotic distributio

More information

ON LOCAL LINEAR ESTIMATION IN NONPARAMETRIC ERRORS-IN-VARIABLES MODELS 1

ON LOCAL LINEAR ESTIMATION IN NONPARAMETRIC ERRORS-IN-VARIABLES MODELS 1 Teory of Stocastic Processes Vol2 28, o3-4, 2006, pp*-* SILVELYN ZWANZIG ON LOCAL LINEAR ESTIMATION IN NONPARAMETRIC ERRORS-IN-VARIABLES MODELS Local liear metods are applied to a oparametric regressio

More information

Uniformly Consistency of the Cauchy-Transformation Kernel Density Estimation Underlying Strong Mixing

Uniformly Consistency of the Cauchy-Transformation Kernel Density Estimation Underlying Strong Mixing Appl. Mat. If. Sci. 7, No. L, 5-9 (203) 5 Applied Matematics & Iformatio Scieces A Iteratioal Joural c 203 NSP Natural Scieces Publisig Cor. Uiformly Cosistecy of te Caucy-Trasformatio Kerel Desity Estimatio

More information

Lecture 7 Testing Nonlinear Inequality Restrictions 1

Lecture 7 Testing Nonlinear Inequality Restrictions 1 Eco 75 Lecture 7 Testig Noliear Iequality Restrictios I Lecture 6, we discussed te testig problems were te ull ypotesis is de ed by oliear equality restrictios: H : ( ) = versus H : ( ) 6= : () We sowed

More information

Lecture 9: Regression: Regressogram and Kernel Regression

Lecture 9: Regression: Regressogram and Kernel Regression STAT 425: Itroductio to Noparametric Statistics Witer 208 Lecture 9: Regressio: Regressogram ad erel Regressio Istructor: Ye-Ci Ce Referece: Capter 5 of All of oparametric statistics 9 Itroductio Let X,

More information

Let us give one more example of MLE. Example 3. The uniform distribution U[0, θ] on the interval [0, θ] has p.d.f.

Let us give one more example of MLE. Example 3. The uniform distribution U[0, θ] on the interval [0, θ] has p.d.f. Lecture 5 Let us give oe more example of MLE. Example 3. The uiform distributio U[0, ] o the iterval [0, ] has p.d.f. { 1 f(x =, 0 x, 0, otherwise The likelihood fuctio ϕ( = f(x i = 1 I(X 1,..., X [0,

More information

Resampling Methods. X (1/2), i.e., Pr (X i m) = 1/2. We order the data: X (1) X (2) X (n). Define the sample median: ( n.

Resampling Methods. X (1/2), i.e., Pr (X i m) = 1/2. We order the data: X (1) X (2) X (n). Define the sample median: ( n. Jauary 1, 2019 Resamplig Methods Motivatio We have so may estimators with the property θ θ d N 0, σ 2 We ca also write θ a N θ, σ 2 /, where a meas approximately distributed as Oce we have a cosistet estimator

More information

Notes On Nonparametric Density Estimation. James L. Powell Department of Economics University of California, Berkeley

Notes On Nonparametric Density Estimation. James L. Powell Department of Economics University of California, Berkeley Notes O Noparametric Desity Estimatio James L Powell Departmet of Ecoomics Uiversity of Califoria, Berkeley Uivariate Desity Estimatio via Numerical Derivatives Cosider te problem of estimatig te desity

More information

Estimation of the essential supremum of a regression function

Estimation of the essential supremum of a regression function Estimatio of the essetial supremum of a regressio fuctio Michael ohler, Adam rzyżak 2, ad Harro Walk 3 Fachbereich Mathematik, Techische Uiversität Darmstadt, Schlossgartestr. 7, 64289 Darmstadt, Germay,

More information

32 estimating the cumulative distribution function

32 estimating the cumulative distribution function 32 estimatig the cumulative distributio fuctio 4.6 types of cofidece itervals/bads Let F be a class of distributio fuctios F ad let θ be some quatity of iterest, such as the mea of F or the whole fuctio

More information

Introductory statistics

Introductory statistics CM9S: Machie Learig for Bioiformatics Lecture - 03/3/06 Itroductory statistics Lecturer: Sriram Sakararama Scribe: Sriram Sakararama We will provide a overview of statistical iferece focussig o the key

More information

ALLOCATING SAMPLE TO STRATA PROPORTIONAL TO AGGREGATE MEASURE OF SIZE WITH BOTH UPPER AND LOWER BOUNDS ON THE NUMBER OF UNITS IN EACH STRATUM

ALLOCATING SAMPLE TO STRATA PROPORTIONAL TO AGGREGATE MEASURE OF SIZE WITH BOTH UPPER AND LOWER BOUNDS ON THE NUMBER OF UNITS IN EACH STRATUM ALLOCATING SAPLE TO STRATA PROPORTIONAL TO AGGREGATE EASURE OF SIZE WIT BOT UPPER AND LOWER BOUNDS ON TE NUBER OF UNITS IN EAC STRATU Lawrece R. Erst ad Cristoper J. Guciardo Erst_L@bls.gov, Guciardo_C@bls.gov

More information

17. Joint distributions of extreme order statistics Lehmann 5.1; Ferguson 15

17. Joint distributions of extreme order statistics Lehmann 5.1; Ferguson 15 17. Joit distributios of extreme order statistics Lehma 5.1; Ferguso 15 I Example 10., we derived the asymptotic distributio of the maximum from a radom sample from a uiform distributio. We did this usig

More information

Kernel density estimator

Kernel density estimator Jauary, 07 NONPARAMETRIC ERNEL DENSITY ESTIMATION I this lecture, we discuss kerel estimatio of probability desity fuctios PDF Noparametric desity estimatio is oe of the cetral problems i statistics I

More information

10/ Statistical Machine Learning Homework #1 Solutions

10/ Statistical Machine Learning Homework #1 Solutions Caregie Mello Uiversity Departet of Statistics & Data Sciece 0/36-70 Statistical Macie Learig Hoework # Solutios Proble [40 pts.] DUE: February, 08 Let X,..., X P were X i [0, ] ad P as desity p. Let p

More information

Sequences and Series of Functions

Sequences and Series of Functions Chapter 6 Sequeces ad Series of Fuctios 6.1. Covergece of a Sequece of Fuctios Poitwise Covergece. Defiitio 6.1. Let, for each N, fuctio f : A R be defied. If, for each x A, the sequece (f (x)) coverges

More information

Efficient GMM LECTURE 12 GMM II

Efficient GMM LECTURE 12 GMM II DECEMBER 1 010 LECTURE 1 II Efficiet The estimator depeds o the choice of the weight matrix A. The efficiet estimator is the oe that has the smallest asymptotic variace amog all estimators defied by differet

More information

Lecture 33: Bootstrap

Lecture 33: Bootstrap Lecture 33: ootstrap Motivatio To evaluate ad compare differet estimators, we eed cosistet estimators of variaces or asymptotic variaces of estimators. This is also importat for hypothesis testig ad cofidece

More information

MASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.436J/15.085J Fall 2008 Lecture 19 11/17/2008 LAWS OF LARGE NUMBERS II THE STRONG LAW OF LARGE NUMBERS

MASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.436J/15.085J Fall 2008 Lecture 19 11/17/2008 LAWS OF LARGE NUMBERS II THE STRONG LAW OF LARGE NUMBERS MASSACHUSTTS INSTITUT OF TCHNOLOGY 6.436J/5.085J Fall 2008 Lecture 9 /7/2008 LAWS OF LARG NUMBRS II Cotets. The strog law of large umbers 2. The Cheroff boud TH STRONG LAW OF LARG NUMBRS While the weak

More information

Economics 241B Relation to Method of Moments and Maximum Likelihood OLSE as a Maximum Likelihood Estimator

Economics 241B Relation to Method of Moments and Maximum Likelihood OLSE as a Maximum Likelihood Estimator Ecoomics 24B Relatio to Method of Momets ad Maximum Likelihood OLSE as a Maximum Likelihood Estimator Uder Assumptio 5 we have speci ed the distributio of the error, so we ca estimate the model parameters

More information

Random Variables, Sampling and Estimation

Random Variables, Sampling and Estimation Chapter 1 Radom Variables, Samplig ad Estimatio 1.1 Itroductio This chapter will cover the most importat basic statistical theory you eed i order to uderstad the ecoometric material that will be comig

More information

Appendix to: Hypothesis Testing for Multiple Mean and Correlation Curves with Functional Data

Appendix to: Hypothesis Testing for Multiple Mean and Correlation Curves with Functional Data Appedix to: Hypothesis Testig for Multiple Mea ad Correlatio Curves with Fuctioal Data Ao Yua 1, Hog-Bi Fag 1, Haiou Li 1, Coli O. Wu, Mig T. Ta 1, 1 Departmet of Biostatistics, Bioiformatics ad Biomathematics,

More information

Econometrica Supplementary Material

Econometrica Supplementary Material Ecoometrica Supplemetary Material SUPPLEMENT TO NONPARAMETRIC INSTRUMENTAL VARIABLES ESTIMATION OF A QUANTILE REGRESSION MODEL : MATHEMATICAL APPENDIX (Ecoometrica, Vol. 75, No. 4, July 27, 1191 128) BY

More information

Model-based Variance Estimation for Systematic Sampling

Model-based Variance Estimation for Systematic Sampling ASA Sectio o Survey Researc Metods Model-based Variace Estimatio for Systematic Samplig Xiaoxi Li ad J D Opsomer Ceter for Survey Statistics ad Metodology, ad Departmet of Statistics, Iowa State Uiversity,

More information

Estimation in Threshold Autoregressive Models with Nonstationarity

Estimation in Threshold Autoregressive Models with Nonstationarity Te Uiversity of Adelaide Scool of Ecoomics Researc Paper No. 2009-25 Estimatio i Tresold Autoregressive Models wit Nostatioarity Jiti Gao, Dag Tjøsteim ad Jiyig Yi Te Uiversity of Adelaide, Scool of Ecoomics

More information

Week 6. Intuition: Let p (x, y) be the joint density of (X, Y ) and p (x) be the marginal density of X. Then. h dy =

Week 6. Intuition: Let p (x, y) be the joint density of (X, Y ) and p (x) be the marginal density of X. Then. h dy = Week 6 Lecture Kerel regressio ad miimax rates Model: Observe Y i = f (X i ) + ξ i were ξ i, i =,,,, iid wit Eξ i = 0 We ofte assume X i are iid or X i = i/ Nadaraya Watso estimator Let w i (x) = K ( X

More information

LECTURE 8: ASYMPTOTICS I

LECTURE 8: ASYMPTOTICS I LECTURE 8: ASYMPTOTICS I We are iterested i the properties of estimators as. Cosider a sequece of radom variables {, X 1}. N. M. Kiefer, Corell Uiversity, Ecoomics 60 1 Defiitio: (Weak covergece) A sequece

More information

It should be unbiased, or approximately unbiased. Variance of the variance estimator should be small. That is, the variance estimator is stable.

It should be unbiased, or approximately unbiased. Variance of the variance estimator should be small. That is, the variance estimator is stable. Chapter 10 Variace Estimatio 10.1 Itroductio Variace estimatio is a importat practical problem i survey samplig. Variace estimates are used i two purposes. Oe is the aalytic purpose such as costructig

More information

7.1 Convergence of sequences of random variables

7.1 Convergence of sequences of random variables Chapter 7 Limit Theorems Throughout this sectio we will assume a probability space (, F, P), i which is defied a ifiite sequece of radom variables (X ) ad a radom variable X. The fact that for every ifiite

More information

Web-based Supplementary Materials for A Modified Partial Likelihood Score Method for Cox Regression with Covariate Error Under the Internal

Web-based Supplementary Materials for A Modified Partial Likelihood Score Method for Cox Regression with Covariate Error Under the Internal Web-based Supplemetary Materials for A Modified Partial Likelihood Score Method for Cox Regressio with Covariate Error Uder the Iteral Validatio Desig by David M. Zucker, Xi Zhou, Xiaomei Liao, Yi Li,

More information

ECONOMETRIC THEORY. MODULE XIII Lecture - 34 Asymptotic Theory and Stochastic Regressors

ECONOMETRIC THEORY. MODULE XIII Lecture - 34 Asymptotic Theory and Stochastic Regressors ECONOMETRIC THEORY MODULE XIII Lecture - 34 Asymptotic Theory ad Stochastic Regressors Dr. Shalabh Departmet of Mathematics ad Statistics Idia Istitute of Techology Kapur Asymptotic theory The asymptotic

More information

STA Object Data Analysis - A List of Projects. January 18, 2018

STA Object Data Analysis - A List of Projects. January 18, 2018 STA 6557 Jauary 8, 208 Object Data Aalysis - A List of Projects. Schoeberg Mea glaucomatous shape chages of the Optic Nerve Head regio i aimal models 2. Aalysis of VW- Kedall ati-mea shapes with a applicatio

More information

Convergence of random variables. (telegram style notes) P.J.C. Spreij

Convergence of random variables. (telegram style notes) P.J.C. Spreij Covergece of radom variables (telegram style otes).j.c. Spreij this versio: September 6, 2005 Itroductio As we kow, radom variables are by defiitio measurable fuctios o some uderlyig measurable space

More information

Bias reduction in local polynomial regression estimation via global Lipschitz conditions

Bias reduction in local polynomial regression estimation via global Lipschitz conditions Bias reductio i local polyomial regressio estimatio via global Lipscitz coditios Na Kyeog Lee Departmet of Ecoomics Te Uiversity of Colorado Boulder, CO 839-256 USA email: a.lee@colorado.edu Voice: + 62

More information

A Note on the Kolmogorov-Feller Weak Law of Large Numbers

A Note on the Kolmogorov-Feller Weak Law of Large Numbers Joural of Mathematical Research with Applicatios Mar., 015, Vol. 35, No., pp. 3 8 DOI:10.3770/j.iss:095-651.015.0.013 Http://jmre.dlut.edu.c A Note o the Kolmogorov-Feller Weak Law of Large Numbers Yachu

More information

7.1 Convergence of sequences of random variables

7.1 Convergence of sequences of random variables Chapter 7 Limit theorems Throughout this sectio we will assume a probability space (Ω, F, P), i which is defied a ifiite sequece of radom variables (X ) ad a radom variable X. The fact that for every ifiite

More information

Lecture 6 Simple alternatives and the Neyman-Pearson lemma

Lecture 6 Simple alternatives and the Neyman-Pearson lemma STATS 00: Itroductio to Statistical Iferece Autum 06 Lecture 6 Simple alteratives ad the Neyma-Pearso lemma Last lecture, we discussed a umber of ways to costruct test statistics for testig a simple ull

More information

Estimating the Population Mean using Stratified Double Ranked Set Sample

Estimating the Population Mean using Stratified Double Ranked Set Sample Estimatig te Populatio Mea usig Stratified Double Raked Set Sample Mamoud Syam * Kamarulzama Ibraim Amer Ibraim Al-Omari Qatar Uiversity Foudatio Program Departmet of Mat ad Computer P.O.Box (7) Doa State

More information

Computation Of Asymptotic Distribution For Semiparametric GMM Estimators

Computation Of Asymptotic Distribution For Semiparametric GMM Estimators Computatio Of Asymptotic Distributio For Semiparametric GMM Estimators Hideiko Icimura Departmet of Ecoomics Uiversity College Lodo Cemmap UCL ad IFS April 9, 2004 Abstract A set of su ciet coditios for

More information

Mathematical Statistics - MS

Mathematical Statistics - MS Paper Specific Istructios. The examiatio is of hours duratio. There are a total of 60 questios carryig 00 marks. The etire paper is divided ito three sectios, A, B ad C. All sectios are compulsory. Questios

More information

Summary. Recap ... Last Lecture. Summary. Theorem

Summary. Recap ... Last Lecture. Summary. Theorem Last Lecture Biostatistics 602 - Statistical Iferece Lecture 23 Hyu Mi Kag April 11th, 2013 What is p-value? What is the advatage of p-value compared to hypothesis testig procedure with size α? How ca

More information

Semiparametric Estimation in Multivariate Nonstationary Time Series Models

Semiparametric Estimation in Multivariate Nonstationary Time Series Models ISSN 440-77X Australia Departmet of Ecoometrics ad Busiess Statistics ttp://www.buseco.moas.edu.au/depts/ebs/pubs/wpapers/ Semiparametric Estimatio i Multivariate Nostatioary Time Series Models Jiti Gao

More information

UNIFORM CONVERGENCE RATES FOR KERNEL ESTIMATION WITH DEPENDENT DATA

UNIFORM CONVERGENCE RATES FOR KERNEL ESTIMATION WITH DEPENDENT DATA Ecoometric Teory, 24, 2008, 726 748+ Prited i te Uited States of America+ doi: 10+10170S0266466608080304 UNIFORM CONVERGENCE RATES FOR KERNEL ESTIMATION WITH DEPENDENT DATA BRUCE E. HANSEN Uiversity of

More information

Rank tests and regression rank scores tests in measurement error models

Rank tests and regression rank scores tests in measurement error models Rak tests ad regressio rak scores tests i measuremet error models J. Jurečková ad A.K.Md.E. Saleh Charles Uiversity i Prague ad Carleto Uiversity i Ottawa Abstract The rak ad regressio rak score tests

More information

Lecture 19: Convergence

Lecture 19: Convergence Lecture 19: Covergece Asymptotic approach I statistical aalysis or iferece, a key to the success of fidig a good procedure is beig able to fid some momets ad/or distributios of various statistics. I may

More information

Goodness-Of-Fit For The Generalized Exponential Distribution. Abstract

Goodness-Of-Fit For The Generalized Exponential Distribution. Abstract Goodess-Of-Fit For The Geeralized Expoetial Distributio By Amal S. Hassa stitute of Statistical Studies & Research Cairo Uiversity Abstract Recetly a ew distributio called geeralized expoetial or expoetiated

More information

Probability 2 - Notes 10. Lemma. If X is a random variable and g(x) 0 for all x in the support of f X, then P(g(X) 1) E[g(X)].

Probability 2 - Notes 10. Lemma. If X is a random variable and g(x) 0 for all x in the support of f X, then P(g(X) 1) E[g(X)]. Probability 2 - Notes 0 Some Useful Iequalities. Lemma. If X is a radom variable ad g(x 0 for all x i the support of f X, the P(g(X E[g(X]. Proof. (cotiuous case P(g(X Corollaries x:g(x f X (xdx x:g(x

More information

Chapter 3. Strong convergence. 3.1 Definition of almost sure convergence

Chapter 3. Strong convergence. 3.1 Definition of almost sure convergence Chapter 3 Strog covergece As poited out i the Chapter 2, there are multiple ways to defie the otio of covergece of a sequece of radom variables. That chapter defied covergece i probability, covergece i

More information

of the matrix is =-85, so it is not positive definite. Thus, the first

of the matrix is =-85, so it is not positive definite. Thus, the first BOSTON COLLEGE Departmet of Ecoomics EC771: Ecoometrics Sprig 4 Prof. Baum, Ms. Uysal Solutio Key for Problem Set 1 1. Are the followig quadratic forms positive for all values of x? (a) y = x 1 8x 1 x

More information

5. Likelihood Ratio Tests

5. Likelihood Ratio Tests 1 of 5 7/29/2009 3:16 PM Virtual Laboratories > 9. Hy pothesis Testig > 1 2 3 4 5 6 7 5. Likelihood Ratio Tests Prelimiaries As usual, our startig poit is a radom experimet with a uderlyig sample space,

More information

62. Power series Definition 16. (Power series) Given a sequence {c n }, the series. c n x n = c 0 + c 1 x + c 2 x 2 + c 3 x 3 +

62. Power series Definition 16. (Power series) Given a sequence {c n }, the series. c n x n = c 0 + c 1 x + c 2 x 2 + c 3 x 3 + 62. Power series Defiitio 16. (Power series) Give a sequece {c }, the series c x = c 0 + c 1 x + c 2 x 2 + c 3 x 3 + is called a power series i the variable x. The umbers c are called the coefficiets of

More information

Probability and Statistics

Probability and Statistics ICME Refresher Course: robability ad Statistics Staford Uiversity robability ad Statistics Luyag Che September 20, 2016 1 Basic robability Theory 11 robability Spaces A probability space is a triple (Ω,

More information

8.1 Introduction. 8. Nonparametric Inference Using Orthogonal Functions

8.1 Introduction. 8. Nonparametric Inference Using Orthogonal Functions 8. Noparametric Iferece Usig Orthogoal Fuctios 1. Itroductio. Noparametric Regressio 3. Irregular Desigs 4. Desity Estimatio 5. Compariso of Methods 8.1 Itroductio Use a orthogoal basis to covert oparametric

More information

Model Specification Testing in Nonparametric and Semiparametric Time Series Econometrics 1

Model Specification Testing in Nonparametric and Semiparametric Time Series Econometrics 1 Model Specification Testing in Nonparametric and Semiparametric Time Series Econometrics 1 By Jiti Gao 2 and Maxwell King 3 Abstract We propose a simultaneous model specification procedure for te conditional

More information

POWER COMPARISON OF EMPIRICAL LIKELIHOOD RATIO TESTS: SMALL SAMPLE PROPERTIES THROUGH MONTE CARLO STUDIES*

POWER COMPARISON OF EMPIRICAL LIKELIHOOD RATIO TESTS: SMALL SAMPLE PROPERTIES THROUGH MONTE CARLO STUDIES* Kobe Uiversity Ecoomic Review 50(2004) 3 POWER COMPARISON OF EMPIRICAL LIKELIHOOD RATIO TESTS: SMALL SAMPLE PROPERTIES THROUGH MONTE CARLO STUDIES* By HISASHI TANIZAKI There are various kids of oparametric

More information

Regression with an Evaporating Logarithmic Trend

Regression with an Evaporating Logarithmic Trend Regressio with a Evaporatig Logarithmic Tred Peter C. B. Phillips Cowles Foudatio, Yale Uiversity, Uiversity of Aucklad & Uiversity of York ad Yixiao Su Departmet of Ecoomics Yale Uiversity October 5,

More information

ECE 901 Lecture 12: Complexity Regularization and the Squared Loss

ECE 901 Lecture 12: Complexity Regularization and the Squared Loss ECE 90 Lecture : Complexity Regularizatio ad the Squared Loss R. Nowak 5/7/009 I the previous lectures we made use of the Cheroff/Hoeffdig bouds for our aalysis of classifier errors. Hoeffdig s iequality

More information

Precise Rates in Complete Moment Convergence for Negatively Associated Sequences

Precise Rates in Complete Moment Convergence for Negatively Associated Sequences Commuicatios of the Korea Statistical Society 29, Vol. 16, No. 5, 841 849 Precise Rates i Complete Momet Covergece for Negatively Associated Sequeces Dae-Hee Ryu 1,a a Departmet of Computer Sciece, ChugWoo

More information

Last Lecture. Wald Test

Last Lecture. Wald Test Last Lecture Biostatistics 602 - Statistical Iferece Lecture 22 Hyu Mi Kag April 9th, 2013 Is the exact distributio of LRT statistic typically easy to obtai? How about its asymptotic distributio? For testig

More information

Properties and Hypothesis Testing

Properties and Hypothesis Testing Chapter 3 Properties ad Hypothesis Testig 3.1 Types of data The regressio techiques developed i previous chapters ca be applied to three differet kids of data. 1. Cross-sectioal data. 2. Time series data.

More information

Berry-Esseen bounds for self-normalized martingales

Berry-Esseen bounds for self-normalized martingales Berry-Essee bouds for self-ormalized martigales Xiequa Fa a, Qi-Ma Shao b a Ceter for Applied Mathematics, Tiaji Uiversity, Tiaji 30007, Chia b Departmet of Statistics, The Chiese Uiversity of Hog Kog,

More information

A goodness-of-fit test based on the empirical characteristic function and a comparison of tests for normality

A goodness-of-fit test based on the empirical characteristic function and a comparison of tests for normality A goodess-of-fit test based o the empirical characteristic fuctio ad a compariso of tests for ormality J. Marti va Zyl Departmet of Mathematical Statistics ad Actuarial Sciece, Uiversity of the Free State,

More information

1.010 Uncertainty in Engineering Fall 2008

1.010 Uncertainty in Engineering Fall 2008 MIT OpeCourseWare http://ocw.mit.edu.00 Ucertaity i Egieerig Fall 2008 For iformatio about citig these materials or our Terms of Use, visit: http://ocw.mit.edu.terms. .00 - Brief Notes # 9 Poit ad Iterval

More information

Kolmogorov-Smirnov type Tests for Local Gaussianity in High-Frequency Data

Kolmogorov-Smirnov type Tests for Local Gaussianity in High-Frequency Data Proceedigs 59th ISI World Statistics Cogress, 5-30 August 013, Hog Kog (Sessio STS046) p.09 Kolmogorov-Smirov type Tests for Local Gaussiaity i High-Frequecy Data George Tauche, Duke Uiversity Viktor Todorov,

More information

EECS564 Estimation, Filtering, and Detection Hwk 2 Solns. Winter p θ (z) = (2θz + 1 θ), 0 z 1

EECS564 Estimation, Filtering, and Detection Hwk 2 Solns. Winter p θ (z) = (2θz + 1 θ), 0 z 1 EECS564 Estimatio, Filterig, ad Detectio Hwk 2 Sols. Witer 25 4. Let Z be a sigle observatio havig desity fuctio where. p (z) = (2z + ), z (a) Assumig that is a oradom parameter, fid ad plot the maximum

More information

Distribution of Random Samples & Limit theorems

Distribution of Random Samples & Limit theorems STAT/MATH 395 A - PROBABILITY II UW Witer Quarter 2017 Néhémy Lim Distributio of Radom Samples & Limit theorems 1 Distributio of i.i.d. Samples Motivatig example. Assume that the goal of a study is to

More information

A RANK STATISTIC FOR NON-PARAMETRIC K-SAMPLE AND CHANGE POINT PROBLEMS

A RANK STATISTIC FOR NON-PARAMETRIC K-SAMPLE AND CHANGE POINT PROBLEMS J. Japa Statist. Soc. Vol. 41 No. 1 2011 67 73 A RANK STATISTIC FOR NON-PARAMETRIC K-SAMPLE AND CHANGE POINT PROBLEMS Yoichi Nishiyama* We cosider k-sample ad chage poit problems for idepedet data i a

More information

Estimation of the Mean and the ACVF

Estimation of the Mean and the ACVF Chapter 5 Estimatio of the Mea ad the ACVF A statioary process {X t } is characterized by its mea ad its autocovariace fuctio γ ), ad so by the autocorrelatio fuctio ρ ) I this chapter we preset the estimators

More information

TAMS24: Notations and Formulas

TAMS24: Notations and Formulas TAMS4: Notatios ad Formulas Basic otatios ad defiitios X: radom variable stokastiska variabel Mea Vätevärde: µ = X = by Xiagfeg Yag kpx k, if X is discrete, xf Xxdx, if X is cotiuous Variace Varias: =

More information

REGRESSION WITH QUADRATIC LOSS

REGRESSION WITH QUADRATIC LOSS REGRESSION WITH QUADRATIC LOSS MAXIM RAGINSKY Regressio with quadratic loss is aother basic problem studied i statistical learig theory. We have a radom couple Z = X, Y ), where, as before, X is a R d

More information

Entropy Rates and Asymptotic Equipartition

Entropy Rates and Asymptotic Equipartition Chapter 29 Etropy Rates ad Asymptotic Equipartitio Sectio 29. itroduces the etropy rate the asymptotic etropy per time-step of a stochastic process ad shows that it is well-defied; ad similarly for iformatio,

More information

Econ 325 Notes on Point Estimator and Confidence Interval 1 By Hiro Kasahara

Econ 325 Notes on Point Estimator and Confidence Interval 1 By Hiro Kasahara Poit Estimator Eco 325 Notes o Poit Estimator ad Cofidece Iterval 1 By Hiro Kasahara Parameter, Estimator, ad Estimate The ormal probability desity fuctio is fully characterized by two costats: populatio

More information

MASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.265/15.070J Fall 2013 Lecture 3 9/11/2013. Large deviations Theory. Cramér s Theorem

MASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.265/15.070J Fall 2013 Lecture 3 9/11/2013. Large deviations Theory. Cramér s Theorem MASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.265/5.070J Fall 203 Lecture 3 9//203 Large deviatios Theory. Cramér s Theorem Cotet.. Cramér s Theorem. 2. Rate fuctio ad properties. 3. Chage of measure techique.

More information

Stat 200 -Testing Summary Page 1

Stat 200 -Testing Summary Page 1 Stat 00 -Testig Summary Page 1 Mathematicias are like Frechme; whatever you say to them, they traslate it ito their ow laguage ad forthwith it is somethig etirely differet Goethe 1 Large Sample Cofidece

More information

Lecture 7: Properties of Random Samples

Lecture 7: Properties of Random Samples Lecture 7: Properties of Radom Samples 1 Cotiued From Last Class Theorem 1.1. Let X 1, X,...X be a radom sample from a populatio with mea µ ad variace σ

More information

Statistical Inference Based on Extremum Estimators

Statistical Inference Based on Extremum Estimators T. Rotheberg Fall, 2007 Statistical Iferece Based o Extremum Estimators Itroductio Suppose 0, the true value of a p-dimesioal parameter, is kow to lie i some subset S R p : Ofte we choose to estimate 0

More information

Problem Set 4 Due Oct, 12

Problem Set 4 Due Oct, 12 EE226: Radom Processes i Systems Lecturer: Jea C. Walrad Problem Set 4 Due Oct, 12 Fall 06 GSI: Assae Gueye This problem set essetially reviews detectio theory ad hypothesis testig ad some basic otios

More information

Open book and notes. 120 minutes. Cover page and six pages of exam. No calculators.

Open book and notes. 120 minutes. Cover page and six pages of exam. No calculators. IE 330 Seat # Ope book ad otes 120 miutes Cover page ad six pages of exam No calculators Score Fial Exam (example) Schmeiser Ope book ad otes No calculator 120 miutes 1 True or false (for each, 2 poits

More information

[ ] ( ) ( ) [ ] ( ) 1 [ ] [ ] Sums of Random Variables Y = a 1 X 1 + a 2 X 2 + +a n X n The expected value of Y is:

[ ] ( ) ( ) [ ] ( ) 1 [ ] [ ] Sums of Random Variables Y = a 1 X 1 + a 2 X 2 + +a n X n The expected value of Y is: PROBABILITY FUNCTIONS A radom variable X has a probabilit associated with each of its possible values. The probabilit is termed a discrete probabilit if X ca assume ol discrete values, or X = x, x, x 3,,

More information

Study the bias (due to the nite dimensional approximation) and variance of the estimators

Study the bias (due to the nite dimensional approximation) and variance of the estimators 2 Series Methods 2. Geeral Approach A model has parameters (; ) where is ite-dimesioal ad is oparametric. (Sometimes, there is o :) We will focus o regressio. The fuctio is approximated by a series a ite

More information

1 The Haar functions and the Brownian motion

1 The Haar functions and the Brownian motion 1 The Haar fuctios ad the Browia motio 1.1 The Haar fuctios ad their completeess The Haar fuctios The basic Haar fuctio is 1 if x < 1/2, ψx) = 1 if 1/2 x < 1, otherwise. 1.1) It has mea zero 1 ψx)dx =,

More information

This exam contains 19 pages (including this cover page) and 10 questions. A Formulae sheet is provided with the exam.

This exam contains 19 pages (including this cover page) and 10 questions. A Formulae sheet is provided with the exam. Probability ad Statistics FS 07 Secod Sessio Exam 09.0.08 Time Limit: 80 Miutes Name: Studet ID: This exam cotais 9 pages (icludig this cover page) ad 0 questios. A Formulae sheet is provided with the

More information

Tests of Hypotheses Based on a Single Sample (Devore Chapter Eight)

Tests of Hypotheses Based on a Single Sample (Devore Chapter Eight) Tests of Hypotheses Based o a Sigle Sample Devore Chapter Eight MATH-252-01: Probability ad Statistics II Sprig 2018 Cotets 1 Hypothesis Tests illustrated with z-tests 1 1.1 Overview of Hypothesis Testig..........

More information

EXTREMAL PROPERTIES OF HALF-SPACES FOR LOG-CONCAVE DISTRIBUTIONS. BY S. BOBKOV Syktyvkar University

EXTREMAL PROPERTIES OF HALF-SPACES FOR LOG-CONCAVE DISTRIBUTIONS. BY S. BOBKOV Syktyvkar University Te Aals of Probability 1996, Vol. 24, No. 1, 3548 EXTREMAL PROPERTIES O HAL-SPACES OR LOG-CONCAVE DISTRIBUTIONS BY S. BOBKOV Syktyvkar Uiversity Te isoperimetric problem for log-cocave product measures

More information

Linear regression. Daniel Hsu (COMS 4771) (y i x T i β)2 2πσ. 2 2σ 2. 1 n. (x T i β y i ) 2. 1 ˆβ arg min. β R n d

Linear regression. Daniel Hsu (COMS 4771) (y i x T i β)2 2πσ. 2 2σ 2. 1 n. (x T i β y i ) 2. 1 ˆβ arg min. β R n d Liear regressio Daiel Hsu (COMS 477) Maximum likelihood estimatio Oe of the simplest liear regressio models is the followig: (X, Y ),..., (X, Y ), (X, Y ) are iid radom pairs takig values i R d R, ad Y

More information

x x x 2x x N ( ) p NUMERICAL METHODS UNIT-I-SOLUTION OF EQUATIONS AND EIGENVALUE PROBLEMS By Newton-Raphson formula

x x x 2x x N ( ) p NUMERICAL METHODS UNIT-I-SOLUTION OF EQUATIONS AND EIGENVALUE PROBLEMS By Newton-Raphson formula NUMERICAL METHODS UNIT-I-SOLUTION OF EQUATIONS AND EIGENVALUE PROBLEMS. If g( is cotiuous i [a,b], te uder wat coditio te iterative (or iteratio metod = g( as a uique solutio i [a,b]? '( i [a,b].. Wat

More information

On the convergence, consistence and stability of a standard finite difference scheme

On the convergence, consistence and stability of a standard finite difference scheme AMERICAN JOURNAL OF SCIENTIFIC AND INDUSTRIAL RESEARCH 2, Sciece Huβ, ttp://www.sciub.org/ajsir ISSN: 253-649X, doi:.525/ajsir.2.2.2.74.78 O te covergece, cosistece ad stabilit of a stadard fiite differece

More information

Math Solutions to homework 6

Math Solutions to homework 6 Math 175 - Solutios to homework 6 Cédric De Groote November 16, 2017 Problem 1 (8.11 i the book): Let K be a compact Hermitia operator o a Hilbert space H ad let the kerel of K be {0}. Show that there

More information

STAT Homework 1 - Solutions

STAT Homework 1 - Solutions STAT-36700 Homework 1 - Solutios Fall 018 September 11, 018 This cotais solutios for Homework 1. Please ote that we have icluded several additioal commets ad approaches to the problems to give you better

More information

MASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.265/15.070J Fall 2013 Lecture 21 11/27/2013

MASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.265/15.070J Fall 2013 Lecture 21 11/27/2013 MASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.265/15.070J Fall 2013 Lecture 21 11/27/2013 Fuctioal Law of Large Numbers. Costructio of the Wieer Measure Cotet. 1. Additioal techical results o weak covergece

More information

Stability analysis of numerical methods for stochastic systems with additive noise

Stability analysis of numerical methods for stochastic systems with additive noise Stability aalysis of umerical metods for stoctic systems wit additive oise Yosiiro SAITO Abstract Stoctic differetial equatios (SDEs) represet pysical peomea domiated by stoctic processes As for determiistic

More information

Lecture Notes 15 Hypothesis Testing (Chapter 10)

Lecture Notes 15 Hypothesis Testing (Chapter 10) 1 Itroductio Lecture Notes 15 Hypothesis Testig Chapter 10) Let X 1,..., X p θ x). Suppose we we wat to kow if θ = θ 0 or ot, where θ 0 is a specific value of θ. For example, if we are flippig a coi, we

More information

A proposed discrete distribution for the statistical modeling of

A proposed discrete distribution for the statistical modeling of It. Statistical Ist.: Proc. 58th World Statistical Cogress, 0, Dubli (Sessio CPS047) p.5059 A proposed discrete distributio for the statistical modelig of Likert data Kidd, Marti Cetre for Statistical

More information

FACULTY OF MATHEMATICAL STUDIES MATHEMATICS FOR PART I ENGINEERING. Lectures

FACULTY OF MATHEMATICAL STUDIES MATHEMATICS FOR PART I ENGINEERING. Lectures FACULTY OF MATHEMATICAL STUDIES MATHEMATICS FOR PART I ENGINEERING Lectures MODULE 5 STATISTICS II. Mea ad stadard error of sample data. Biomial distributio. Normal distributio 4. Samplig 5. Cofidece itervals

More information

MASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.265/15.070J Fall 2013 Lecture 6 9/23/2013. Brownian motion. Introduction

MASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.265/15.070J Fall 2013 Lecture 6 9/23/2013. Brownian motion. Introduction MASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.265/5.070J Fall 203 Lecture 6 9/23/203 Browia motio. Itroductio Cotet.. A heuristic costructio of a Browia motio from a radom walk. 2. Defiitio ad basic properties

More information