* University of Amsterdam, The Netherlands. This paper was written

Size: px
Start display at page:

Download "* University of Amsterdam, The Netherlands. This paper was written"

Transcription

1 A TEST FOR MODEL SPECIFICATION IN THE ABSENCE OF ALTERNATIVE HYPOTHESIS ", by Herman J. Bierens * Discussin Paper N , Octber 1981 * University f Amsterdam, The Netherlands. This paper was written while I was spending a three-mnth term at the University f Minnesta, whse ind hspitality is gratefully acnwledged. This visit is supprted by a grant frm the Netherlands Organisatin fr the Advancement f Pure Science (Z.W.O.). Center fr Ecnmic Research Department f Ecnmics University f Minnesta Minneaplis, Minnesta 55455

2 Abstract In recent papers ([6], [4]) several tests have been prpsed fr testing the truth f a nnlinear specificatin f a regr.essin mdel against ne r mre well-specified alternative hypthesis. In this paper we first shw by an example that these tests may be defective if all the hyptheses are false, and then we prpse a new test that des nt need alternative hyptheses in the sense that we nly test whether r nt a given specificatin is true. Crrespndence shuld be addressed t: Herman J. Bierens Stichting vr Ecnmisch Onderze der Universiteit van Amsterdam Jdenbreestraat NH Amsterdam The Netherlands

3 1. INTRODUCTION In recent Ecnmetrica articles Pesaran and Deatn [6] and Davidsn and Mac Kinnn [4] prpse several tests fr testing a (nn) linear specificatin f a regressin mdel against ne r mre alternatives. These tests seem t perfrm well if either the null hypthesis r ne f the alternative hyptheses is true. Hwever, althugh these tests are able t reject all the hyptheses, it is pssible that if all the hyptheses are false the least wrst hypthesis will be accepted with relative high prbability. We shw this by applying the P-test f Davidsn and Mac Kinnn [4] t the fllwing example. Cnsider an independent sample [(xl,l' x 2,1' 1)'., (xl,n' x 2,n' en)} frm the trivariate standard nrmal distributin and let fr j = 1,2,.,n, Suppse we have specified this relatinship as where and In rder t test the truth f this hypthesis by the P - test prpsed in [4 J we need a nnnested alternative hypthesis, "thugh nt ne in which we need have any faith" (quting (4), fr example where is as befre and Clearly bth hyptheses are false, but H is clser t the truth than H l. Fllwing [4], we dente:

4 ft A.,. (f(x 1, ~),..., AT A g.,. (g(x p y), I.. M = I - X(X T X)-l XT, 0 T Y =..., (Y1 ' Y n ), f (X, ~» n g(x n, y» where and A yare the OLS estimates f ~ and y, respectively, and X is the 2 X n matrix with elements x.. (i = 1, 2; j = 1, 2, 1.,J..., n) The test statistic f the P - test is (1.1) where (1.2) ~ T.... (g - r) M (y - f) and " (1.3) A,.,.,. A 2 a = n 1 1\ y - f - Ct p M (g - f) II As is shwn in [4] the as N(O, 1) if H is true.,. test statistic t p H nr H1 is true, but nevertheless we have is asympttically distributed In the case under review, hwever, neither (1.4) t... N(O, 4) in distributin as n... c, p as will be prved in Appendix 1. Thus if we prceed the test at the five percent Significance level the prbability that we accept the false null is 1 2 -'2 u du.,. J e ~ du~0.673

5 - 3 - This example shws that the P - test prpsed by Davidsn and Mac Kinnn is nt apprpriate if there is a pssibility that nne f the hyptheses invlved is true. The same applies t their J - test, because fr the example under review the J - test and the P - test yield identical test statistics. We have nt checed thrugh the perfrmance f the test prpsed by Pesaran and Deatn [6J fr the abve case, but it seems t be liely that we wuld find similar results. Our argument maes clear that the prblem f testing the functinal. frm f a regressin mdel is nt yet cnclusively slved, hence there is still a need fr an alternative specificatin test that is mre watertight than thse we just have discussed. In this paper we shall prpse such a-'". r. test. This test has the advantage that it des nt need any specified t alternative hypthesis; we nly test whether r nt a given specificatin is true. In sectin 2 we shall give a review f this test prcedure. In sectin 3 we shall deal with the asympttic thery.

6 IDENTIFICATION OF THE NULL HYPOTHESIS AND SUMMARY OF!HE TEST PROCEDURE The Hypthesis Cnsider an i.i.d. stchastic prcess (Yl' Xl)'.., (Y n ' xn) in It X It, where the satisfy (2.1) The cnditin (2.1) is sufficient fr the existence f a Brel measurable functin g ( ) n It such that (2.2) (See Chung [3, Therem 9.1.2]. Defining u j = Y j - g(x j ) we then get the fllwing tautlgical regressin mael (2.3) j = 1, 2,..., n, where bviusly the u. J satisfy (2.4) a. s., j = 1,2,..., n,... Suppse we have specified the regressin functin g(x) as f (x, e ), where f(x, e) is a nwn real valued Brel measurable functin n R X 9 and e is a parameter space cntaining the unnwn parameter e if this specificatin is true. Testing the truth f this specificatinwe thus have t test the null hypthesis (2.5) against the alternative hypthesis that H fr sme is false: e e e (2.6) fr all e e e

7 - 5 - Identificatin But hw can we identify H versus Hl frm the distributin f the data, r with ther wrds, hw d H and H l, respectively, crrespnd with distinct characteristics f the distributin f (Yj' x j )? The answer t this questin is suggested by the fllwing fundamental therem. THEOREM 1. Elvl < a) (I) Let v be a real valued randm variable satisfying and let z be a randm vectr in P[E(v\z) = 0] < 1 if and nly if R Then we have: Eve ittz :F 0 fr sme nnrandm vectr t e 'R. (II) If in additin. z is bunded then p[e(vlz) = 0] < 1 ittz if and nly if Eve 0 ~ 0 fr sme nnrandm vec tr t e'r in an arbitrarily small neighbrhd f t = 0 Prf: Appendix 2. Thus frm part I f therem 1 it fllws that prbabili ty 1 fr sme fr all t e E,, hence - f(x., a)lxjj = 0 with J T it X if and nly if E(Yj - f(x j, 9 0» e J = 0 COROLLARY l! H is true if and nly if fr sme 9 e : T 0 it x. E(y. - f(x.,a» e J - 0 fr all t E'R J J 0 Als part II f therem 1 can be used fr identifying the hyptheses, as we nw shw. Let ~ be a bunded Brel measurable mapping frm R int such that x. J fr example and ~ (x j ) generate the same Euclidean Brel field, atg (Xl.),J (2.7) atg (x.),j

8 - 6 - Then (2.8) hence, applying part II f therem 1, we have: COROLLARY 2: Let ~ be any bunded Brel measurable mapping frm R int such that and x. generate the same Euclidean J Brel field. Then Hl is true if and nly if fr every e e there is a t in an arbitrarily small neighbrhd f t = 0 such that itaf(xj) E (y j - f (x r e)) e '" 0. Of curse, if H is true then fr sme fr all t e R but it is sufficient t verify this nly fr an arbitrarily small neighbrhd f t = O. A direct cnsequence f crllary 2 is: COROLLARY 3: Let ~ be defined as in crllary 2. Let N = X [- e:" e:. 1, t=l.",.", where t > 0 (t = 1, 2,, ) are arbitrarily chsen. Put (2.9) n(e) = IE(Yj - f(x j, itt~(xj)12 e)) e dt 0 Then H is true if.,,(9 0 ) = 0 fr sme e e and Hl is true if inf n(9) > 0. eee The Test This result suggests t test H against Hl by using a test statistic f the frm: (2.10)... where N and ~ are as befre and e is the nnlinear least squares estimatr f e 1.' e 0'

9 - 7 - (2.11) e ee a.s., Nte that if we put ~ 2 n 2 f(x, j e» = inf ~ (y. - f(x., s» see j=l 3 3 (2.12) and (2.13) ~. y. - f(x., e), J 3 3 ~ can be written as (2.14) In sectin 3 we shw that under H,. n 11 cnverges in distributin t a nnnegative randm variable with first mment I.l, say, and that we can estimate this mment cnsistently by (2.15),.,.2 IJ :a 2 a [IT It-t, - "'-1 -A } tr (A.D), -t,=l where,. a is the usual estimate f the variance f the u., 3 (2.16) and (2.17) ~ In n T ~,. B=2 l: ~ [(/a9 )f(x.],e)} [C%S)f(x., S)} n j =1 j = sin[!,(z,.-z,.)] J """,J1 """,J2 II t=l S by applying Chebishev's inequality (fr first mments) we cnclude that fr any Ci g (0, 1),

10 - 8 - (2.18) A 1 A limsupp[n'l1>~~} ~a n-d) if H is true. Mrever, fr every a > 0 we have (2.19) A 1 A lim Pen '11 > -!J.] = 1 n-~ a if Hl is true, as als will be shwn in the next sectin. Thus, prceeding the test at the a' 100 percent cnfidence level we accept H if and we reject H if nt. The result (2.18) is smewhat crude, as it is based n Chebishev's inequality, which is nt a very sharp inequality. It wuld, f curse, be better t base the test directly n the limiting distributin f n'l1, but this limiting distributin appears t have a t cmplicated structure fr that. Hwever, if we assume that the distributin f the u j i. 1. d. 2 N(O,Q' ) and that the sequence Cu,) is independent f the. J sequence (x, ) we can by Mnte Carl simulatin estimate a J 1... sharper lwer bund than the bund ~]J used J.' "" n.. (2 18) We shall discuss this pint t in the next sectin. is The abve test is nly ne example ut f a large class f similar tests that can be derived n the ba"sis f therem 1. Fr example, we als may use a test statistic f the frm A 1 n T it x. 2 1, =ll-!: (Yj - f(x j, e))e J I wet) dt n. 1 J= where wet) is a psitive weight functin, fr example a -variate nrmal density, and fr this test statistic we can, n basis f part I f therem 1, derive similar results as (2.18) and (2.19).

11 ASYMPTOTIC THEORY The Assumptins In this sectin we shall set frth cnditins such that (2.18) and (2.19) are true. The cnditins invlved are just thse fr wea cnsistency and asympttic nrmality f the nnlinear least squares estimatr as can be fund in Jennrich [5] and Bierens [1, 2]. Our first assumptin cncerns the distributin f the data. ASSUMPTION 1: The bservatins (y l' Xl)'..., (y, x ) are Ll.d. randm vectrs in R a > 0 x JR Yj The satisfy EIYjI2+~ < fr n n c sme The i.i.d. assumptin is, f curse, very restrictive and rules ut mst f the ecnmetric applicatins. It is, hwever, merely made fr cnvenience. Using the results in [1] and [2] it is pssible t extend the results in this paper t nnstatinary and nnlinear time series regressins, but this will be ut f the scpe f the present paper. The cnditin n the abslute mments f is strnger than (2. 1), but needed in rder t assure that the errrs u. J defined by (3.1) have finite abslute mments f rder slightly larger than 2. Thus we have (3.2) E U j = cr < c, E I u j I < c fr sme 0 > a. Mrever, we shall limit ur attentin t specificatins f(x, e) satisfying the fllwing cnditins.

12 ASSUMPTION 2: The functin f(x, e) and its first and secnd partial derivatives and i2 = 1, 2,., m), are cntinuus real functins n R X 8l, Tl1here tel is a cnvex cmpact subset f Rm. If H is true then e is 0 an interir pint f tel. This assumptin implies that if H. 0 is true then the true regressin functin g is cntinuus n R, but under H1 g may be any Brel measurable functin n R. Furthermre, f1rwing [lj we shall tmpse sme wea cnditins n the mments f f (x., e) and its first and secnd partia 1 deri va ti ves t e, 1. e., J ASSUMPTION 3: There exists a a > 0 such that: In rder t estimate the true parameter e cnsistently by least squares estimatin it shuld be unique in the sense that it is the nly pint in e such that (3.3 ) = E[g(x.) - f(x., e )}2 = inf E[g(x. - f(x., e)}2 J J 0 eee J J ' prvided H is true. But even if H is false we can define a pint 0 e in tel by the right equality in (3.3), and if such a pint is unique - it can be estimated cnsistently by least squares. Therefre we assume:

13 ASSUMPTION 4: Under H as well as under Hl there exists a unique pint 9* in e such that Of curse, under H we have e... e. Mrever we nte that by 0 * assumptin 4 the alternative hypthesis can be restated as (3.4) Frm the assumptins 1 thrugh 4 and Therem in [1) it nw fllws that the leas't squares estimatr e is wealy cnsistent: (3.5) plim e... 9 under H, ~ 0 0 but using the argument in Sectin 3.1 in [1) it is nt hard t shw " that als (3.6) plim e... 9* under Hl. n-'- Next suppse: ASSUMPTION 5: The matrix (3.7) is psitive definite and 2... (J a.s.. Frm the argument in Sectin in [1) it then fllws that under H (3.8) '- '" 1 n -1 T plim l\~n(e - e) - r-!: uja (/ee )f(x.,e)\i = 0 n_a:» 0 \In j=l J 0 s that by the central limit therem (3.9), " n (9-90) - Nm [0, a A ] in distributin

14 Asympttic Thery under H Befre we cntinue ur argument we shall first intrduce sme additinal ntatin. Thus we dente (3.10) (3.11) (3.12) (3.l3) say, where (3.14) E3.15) and (3.16) ~ ( 1 ) (t).. t (y j j=l ~(2) (t)." 1 ~ [u j + (e - n j"l T (3) 1 n it Zj 1 ~ (t)--e u.(e -(/e)f(xj,e)a- S (Tn n. 1 J 0 0 J= 1 n ittz j 1 1 n ~*(t) = n t ujee - (/e) f(x j, e )A- b(t)}.. Ii E U p.(t) j=l 0 j=l j J z. J is defined by (2.12), ittz. P (t).. e J - (ale) f(x., e )A -1 b(t) J J 0 T 1 n T lt z. b (t).. ii E (0/ 09) f (x ' 9 ) e J j=l j 0 ittz b(t).. E S (t).. E (/091f(x., 9 ) e j J 0 Nte that frm (2.10) and (3.10), (3.17),., = J 1 ~(1) (t) \2 dt Nw put 0,..* (3.18),.. * 12,.,- J I ~ (t) dt N We then have THEOREM 2. then p1im n_(x) I If H is true and if the assumptins 1 thrugh 5 hld,..,..*\ n,.,-n,., =0.

15 Prf: (3.19) - Using the mean value therem we can write T ~(l) 1 n it z. ~ (t) = Ii!: [u. + f(x, e) - f(x, j j e)]e J j=l J. T 1 n,. T T _ l.t z. = ii!: (U - (e - e) (~/~e) f(x, e(t)]e J j=1 j j where Set) is a mean value satisfying (3.20) a.s. fr all t e JR Using Therem in II] it is nt hard t shw (3.21) p1im sup I 'In ~(l\t) - rn~(2) (t) I = 0 n-+"o ten and c.mbining (3.8) and (3.11) we see that (3.22) plim sup Irn~(2) (t) _rn~(3) (t) I = 0 ~ ten Again using Therem in 11] we may cnclude that (3.23) and cnsequently that plimsup Ib (t) - bet) 1= 0 n-+<» ten 0 (3.24) plim sup 1m ~(3)(t) - Vn~* (t) I.. 0 n-+<» ten Cmbining (3.21), (3.22) and (3.24) we btain (3.25) plim sup Ivn~(l) (t) - Vn~*(t) I = 0 Il-+<D ten and applying Lemma in II] we cnclude frm (3.25) that als (3.26) This prves the therem. Q.E.D.

16 Therem 2 implies that under H A n't1 and A n 11 * have the same limiting distributin. But what is the limiting distributin f n ~ *? If we substitute (3.13) in (3.18) we get (3.27) where is the cmplex cnjugate f If we wuld be able t as a prduct f Li.d. randm variables,.* n't1 wuld cnverge in distributin t but it appears imp~sible t split up the integral A invlved in this way. S the limiting distributin f n't1* is prbably f an unnwn type. There are, hwever, tw ways ut f this prblem. A* The first 'way ut is t cmpute the expectatin f n'l1 and t apply Chebishev's inequality. This expectatin is: (3.28) ij. = En'T1* = -,. ln i: Eu. E r P.(t)"j (t) dt = a E r p.(t) p.(t) dt n j=l J i J & J J 0 which is bviusly independent f the sample size n. We then have by Chebishev's inequality (3.29) and.cnsequently by therem 2 A E nn* 1 :=0 a -1J Ct (3.30),. 1 lijnsup pen 'T1 > - ij.] ~ Ct a S if we can find a cnsistent estimate ij.' say, f ij. then (3.31) We can (3.32) limsup n- cnstruct such an n b (t) = 1 i: n j=l estimate ij. as fllws. Put. T T,. ~t z. (010 e ) f (x., 9) e J J

17 and (3.33) T,.. it z., P"j(t) - e J - (alae) f(x j, e)a 6(t), where A is defined by (2.16). Then (3.34) because 1 n,..,.. 1 n 1 sup 1- t ~. (t) p. (t).. - i: ~. (t) 1'. (t) ten n j=l J J n j-l J J ~ sup I b(t)t A- 1 bet) - b(t)t A- 1 bet) I ten in prb. (3.35) p1im A = A, n-+where A is defined in assumptin 5, and (3.36) p1im sup Ib(t) - bet) I = p, n- ten, () as is nt hard t verify by using Therem in [1]. Mrever, (3.37) 1 n p1im sup 1- i: ".(t) p.(t) - E p.(t) "j(t) I = 0 n- ten n j=l J J J as is als easily verified by using Therem in [1]. cmbine (3.34) and (3.37) we then may cnclude S if we (3.38) p1im J 1 ~ p.(t) 6.(t)dt-jE {p.(t) I'j(t)}dt = E j p.(t) Pj(t) dt ~ N n j-1 J J N J N J 0 and hence,..,..2 1 n,.. """A~~ (3.39) ~ = a - i: r p.(t) p.(t) dt n j=l ~ J J is a cnsistent estimatr f IJ.. We leave it t the reader t verify that the estimatr ~ defined by (3.39) can be written as (2.15). S we have prved by nw

18 THEOREM 3. If H is true and if the assumptins 1 thrugh 5 hld then fr every a e (O~ 1), > limsup "" 1 "" P (n T'l > ] ~ a. a n~ This is ur first way ut f the prblem that the limiting distributin "* f n T'l is unnwn. The secnd way is the fllwing. If we mae the additinal assumptin that the sequence (u.) f J disturbances (U j = y - E(Y.lx.» is independent f the sequence j J J f regressrs and mrever that these u.' s are distributed as J then by replacing the Uj's in (3.13) by ther independent randm drawings frm N(O, a 2 ) the resulting integral f the type (3.18) has the same distributin as the riginal ne. Thus if we draw an artificial randm sample (Wl~.., W n } frm the standard nrmal distributin and if we put (3.40) (3.41) 1 n ~*(t) =- E aw.p.(t), n j=l J J Ti*(t) = J 1 ~(t) \2 dt, N then n;* and n~* have the same distributin. Hwever, a and Pj(t) are nt bservable. Therefre we replace a in (3.40) by its estimate ; and Pj(t) by ~j(t) (defined by (3.33». Thus, we put: (3.42) ~ 1 n" A s (t)=;- E aw.f).(t), j=l J J (3.43) Ti**". J I ret) 12 dt. N Nw similar t therem 2 it can be shwn that (3.44) p lim InTi ** - n 11 * I = 0 n-+- and hence tha t n T'l -** and n T'l ""* have the same limi ting dis tribu tin Therefre by cmputing (3.43) fr a large number f artificially drawn

19 randm samples [WI", W n } frm the standard nrmal distributin we can establish a (randm) number Pa - such that a x 100 percent f the Ti** 's are larger than p - We then have apprximately a P[n~ > PaJ. :=::$ a fr large n, prvided H is true. Asympttic Thery under HI We have seen that if HI is true then under the assumptins 1 thrugh 4, plim. e =- e*, where e* is defined by assumptin 4. Therefre, if we put (3.45) ittz ~(l) (t) =-1 t (YJ' - f(x J " e*» e j *. n j=-l then frm (3.10), (3.6) and (3.45) (3.46) plim sup I,(l) (t) - ~.il) (t) I = 0 n-+cd ten and mrever, (3.47) where (3.48) plim sup I ~(;) (t) - ~(t) I.. 0 ll'+'d ten, T (1) ~t z ~*(t) = E ~* (t) = E[g(x,) - f(x., e») e j J J * as is nt hard t verify by applying Therem 2.3,4 in [lj. Thus we have under HI (3.49) plim. sup I~(l)(t) - ~(t)1 = 0 n-+- ten and cnsequently (3.50),. plim T1 = 'T'1* = n-+-

20 But frm crllary 3 it fllws that ~ > 0, hence: THEOREM 4. If li is true and if the assumptins 1 thrugh 4 hld l,. then plim n T1 = (l) n... co Since it is nt hard t shw that als under li l the estimatr ~ cnverges in prbability t a finite number, 7herem 4 implies that (2.19) hlds.

21 Prf f (l.4) In additin t the ntatin intrduced in sectin 1 we put (Al.l) (Al.2) z ~ 3 Xz,l 3 x 2,n (Al.3)./2; d e q,. 1,2, Using the ntatins (Al.l) and (Al.2) and realizing that M f =,. M X a = 0 identica lly we may rewrite (l.l) thrugh (l. 3) as: (Al.4) (Al.5) (A1.6) a,.,.,..", 2. (1/ (n - - 1» H y - XS - 0/ M Z y II p 0 We shall need the fllwing results, which are nt hard t prve: (Al.7) &llm (Al. 8) ~im (lin) (lin) x~ = I, T Z Z,. m6 I = lsi, (Al.9) gl..im (lin) Z~,. m l I = 3I, (Al.1O).. g1..im (lin) XTy,. (i) (Al.ll) gum (lin) ZT y = m 4 (i) = 3(i) and cnsequently

22 (Al.12) (Al.D) These results will nw be used fr shwing that asympttically nrmally distributed with zer mean. is First, bserve that (Al.14) Since EZIu"EXI~.. O and since (~i~) isasumf Li.d. randm vectrs in R4 we have by the central limit therem (Al.IS) where (Al.16) z'l' (l/vn( IU) - N4 (0, a) in distributin, X u 1.. -E n ( (ZIu)I (ZIu ), (ZIu)I (XIu) I (Z~u), (XIu) I m 2 + m 4, 0 m 4 + m 6, 0 m 2 + m , 0, 18, 0, 0, 120, 0, 18, 18, 0, 4, 0, Mrever it fllws frm (Al.8), (Al.9) and (Al.13) that (A.17) AI AI T T -1 ( ) plim (y, - y «lin) Z X) «lin) X X)) ).. 5' 5' - 5' ~ tt+"'d say. Thus frm (Al.14) thrugh (Al.17) it fllws

23 (Al.1S)... T T'" T / 96) (l/,fo.)y Z MU-N(O,~ n~) = N\O, 25 in distributin. Next we bserve that we may write (A 1. 19) s that frm (Al.20) (Al.7), (Al.a), (Al.9) and (Al.l3) we have (1) {m~ m:} 12 plim (l/n)\\m Z y\l = (1, 1) 1 2' -"2' = 25 n-+- m6 m6 Cmbining (A l. 5), (Al.1S) and (Al.20) we nw cnclude: (Al.21) in distributin. Mrever, since nw (Al.22) plim a = 0, we have n-+- p plim '2,2 '"' plim (lin) II y- X 112 = plim n- n- n- T 2 (lin) u u = 1I1z +m 2 = 2 and cmbining this result with (Al.20) we get (Al.23) The desired result (1.4) fllws nw easily frm (Al.4), (Al. 21) and (Al.23) Q. E.D.

24 APPENDIX 2 Prf f Therem 1 Prf f Part I Frm Chung [3, Therem 9.1.2] it fllws that there exists a Brel measurable real functin r, say, n R such that (A2.l) Put E(vlz) = r(z) a. s.. (A2.2) Then bviusly r l and r 2 are nnnegative Brel measurable real functins n R satisfying (A2.3) r = Nw assume fr the mment (A2.4) Then we can define prbability measures field ~ as fllws:!jl and n the Euclidean Brel (A2.5)!Jj(B) = J r/x) d!j(x)/c j, B j = 1, 2, where!j is the prbability measure generated by the randm vectr z and B is an arbitrary Brel set in~. Then we may write: (A2. 6) T ittz tt E ve lt z = E r(z) e = J rex) e l x djj(x) say, where

25 (AZ.7) (j = 1, 2) are the characteristic functins f the prbability measures ~j (j = 1, Z), respectively. T If E ve it z == fr every t e R then it fllws frm (AZ.6) that (AZ.8) t E R Hence, substituting t = 0, we get (AZ.9) s that frm (A2.4), (AZ.8) and (AZ.9) (AZ.10) "l(t) = "2(t) fr every t er But (AZ.10) implies that the prbability measures ~l and ~Z are r equal, Le., (AZ.ll) ~l (B) = ~ (B) fr eve ry Bre 1 se t B. Frm (AZ.S), (AZ.9) and (AZ.ll) we nw btain (AZ.12) S r (x) dll 1 ex) f r z (x) d~ ex} fr every Bre 1 se t B and cnsequently B B (AZ.13) But CA2.l4) S rex) dj.j.(xl=o fr every Brel set B. B Bl = (x e It : rex) > 0 } is a Brel set, and thus: CA2.l5) = I rex) dj.j.(x}, 1

26 which is nly pssible if Bi is a null set with respect t!j.. Similarly we cnclude that the Brel set (A2,.16) B2 = [x e]i : rex) < OJ is a null set with respect t!j., and hence (A2.17) is a null set with respect t!j.. This means that r(z) = O. a.s. Thus T we have prved by nw that if (A2.4) hlds and if E ve it z = 0 fr all t e]i then E (vi z) = 0 a. s.. Hwever, if (A2.4) des nt hld then ur cnclusin still hlds, as is nt hard t prve. This cmpletes the "nly if" part f part I f therem 1. Since the "if" part is trivial, part I f therem 1 is prved by nw. r Q.E.D. Prf f Part II Since nw z is bunded we may write (A2.18) S if Eve. T <0 EveJ.tz=Ev1: j=o +0 fr sme i j (ttz)j = CD i j E v (ttz)j j.' 1:., j=o J. t* e:lt then there exists a nnnegative integer j* such that CA2.19) ASSuming that j* is minimal, we therefre have (A2.20) which implies that Eve',., T ].",t* z :f 0 fr an arbitrarily $mall A > 0, say this prves part II f the therem. Q.E.D.

27 REFERENCES [1] BIERENS, H. J.: Rbust Methds and Asympttic Thery in Nnlinear Ecnmetrics, Lecture Ntes in Ecnmics and Mathematical Systems, vl. 192, Heidelberg: Springer-Verlag, [2] BIERENS, H. J.: "A Unifrm Wea Law f Large Numbers under cp - Mixing with Applicatin t Nnlinear Least Squares Estimatin," Statistica Neerlandica, 36 (1982) (t appear). [3] CHUNG, K. L.: A Curse in Prbability Thery, New Yr: Academic Press, [4] DA VInSON, R. AND L. G. MACKINNON: "Several Tests fr Mdel Specificatin in the Presence f Alternative Hyptheses," Ecnmetrica, 49 (1981), r [5] JENNRICH, R. I:: "Asympttic Prperties f Nnlinear Least, Squares Estimatrs;'The Annals f Mathematical Statistics, 40 (1969), [6] PESARAN, M. H. AND A. S. DEATON: IrTesting Nn-nested Nnlinear Regressin Mdels," Ecnmetrica, 46 (1978),

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

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

More information

Internal vs. external validity. External validity. This section is based on Stock and Watson s Chapter 9.

Internal vs. external validity. External validity. This section is based on Stock and Watson s Chapter 9. Sectin 7 Mdel Assessment This sectin is based n Stck and Watsn s Chapter 9. Internal vs. external validity Internal validity refers t whether the analysis is valid fr the ppulatin and sample being studied.

More information

Admissibility Conditions and Asymptotic Behavior of Strongly Regular Graphs

Admissibility Conditions and Asymptotic Behavior of Strongly Regular Graphs Admissibility Cnditins and Asympttic Behavir f Strngly Regular Graphs VASCO MOÇO MANO Department f Mathematics University f Prt Oprt PORTUGAL vascmcman@gmailcm LUÍS ANTÓNIO DE ALMEIDA VIEIRA Department

More information

Lyapunov Stability Stability of Equilibrium Points

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

More information

Homology groups of disks with holes

Homology groups of disks with holes Hmlgy grups f disks with hles THEOREM. Let p 1,, p k } be a sequence f distinct pints in the interir unit disk D n where n 2, and suppse that fr all j the sets E j Int D n are clsed, pairwise disjint subdisks.

More information

Computational modeling techniques

Computational modeling techniques Cmputatinal mdeling techniques Lecture 4: Mdel checing fr ODE mdels In Petre Department f IT, Åb Aademi http://www.users.ab.fi/ipetre/cmpmd/ Cntent Stichimetric matrix Calculating the mass cnservatin relatins

More information

IN a recent article, Geary [1972] discussed the merit of taking first differences

IN a recent article, Geary [1972] discussed the merit of taking first differences The Efficiency f Taking First Differences in Regressin Analysis: A Nte J. A. TILLMAN IN a recent article, Geary [1972] discussed the merit f taking first differences t deal with the prblems that trends

More information

Distributions, spatial statistics and a Bayesian perspective

Distributions, spatial statistics and a Bayesian perspective Distributins, spatial statistics and a Bayesian perspective Dug Nychka Natinal Center fr Atmspheric Research Distributins and densities Cnditinal distributins and Bayes Thm Bivariate nrmal Spatial statistics

More information

CHAPTER 4 DIAGNOSTICS FOR INFLUENTIAL OBSERVATIONS

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

More information

Perturbation approach applied to the asymptotic study of random operators.

Perturbation approach applied to the asymptotic study of random operators. Perturbatin apprach applied t the asympttic study f rm peratrs. André MAS, udvic MENNETEAU y Abstract We prve that, fr the main mdes f stchastic cnvergence (law f large numbers, CT, deviatins principles,

More information

A Matrix Representation of Panel Data

A Matrix Representation of Panel Data web Extensin 6 Appendix 6.A A Matrix Representatin f Panel Data Panel data mdels cme in tw brad varieties, distinct intercept DGPs and errr cmpnent DGPs. his appendix presents matrix algebra representatins

More information

Chapter Summary. Mathematical Induction Strong Induction Recursive Definitions Structural Induction Recursive Algorithms

Chapter Summary. Mathematical Induction Strong Induction Recursive Definitions Structural Induction Recursive Algorithms Chapter 5 1 Chapter Summary Mathematical Inductin Strng Inductin Recursive Definitins Structural Inductin Recursive Algrithms Sectin 5.1 3 Sectin Summary Mathematical Inductin Examples f Prf by Mathematical

More information

Revisiting the Socrates Example

Revisiting the Socrates Example Sectin 1.6 Sectin Summary Valid Arguments Inference Rules fr Prpsitinal Lgic Using Rules f Inference t Build Arguments Rules f Inference fr Quantified Statements Building Arguments fr Quantified Statements

More information

CHAPTER 24: INFERENCE IN REGRESSION. Chapter 24: Make inferences about the population from which the sample data came.

CHAPTER 24: INFERENCE IN REGRESSION. Chapter 24: Make inferences about the population from which the sample data came. MATH 1342 Ch. 24 April 25 and 27, 2013 Page 1 f 5 CHAPTER 24: INFERENCE IN REGRESSION Chapters 4 and 5: Relatinships between tw quantitative variables. Be able t Make a graph (scatterplt) Summarize the

More information

A proposition is a statement that can be either true (T) or false (F), (but not both).

A proposition is a statement that can be either true (T) or false (F), (but not both). 400 lecture nte #1 [Ch 2, 3] Lgic and Prfs 1.1 Prpsitins (Prpsitinal Lgic) A prpsitin is a statement that can be either true (T) r false (F), (but nt bth). "The earth is flat." -- F "March has 31 days."

More information

Resampling Methods. Chapter 5. Chapter 5 1 / 52

Resampling Methods. Chapter 5. Chapter 5 1 / 52 Resampling Methds Chapter 5 Chapter 5 1 / 52 1 51 Validatin set apprach 2 52 Crss validatin 3 53 Btstrap Chapter 5 2 / 52 Abut Resampling An imprtant statistical tl Pretending the data as ppulatin and

More information

V. Balakrishnan and S. Boyd. (To Appear in Systems and Control Letters, 1992) Abstract

V. Balakrishnan and S. Boyd. (To Appear in Systems and Control Letters, 1992) Abstract On Cmputing the WrstCase Peak Gain f Linear Systems V Balakrishnan and S Byd (T Appear in Systems and Cntrl Letters, 99) Abstract Based n the bunds due t Dyle and Byd, we present simple upper and lwer

More information

MATCHING TECHNIQUES. Technical Track Session VI. Emanuela Galasso. The World Bank

MATCHING TECHNIQUES. Technical Track Session VI. Emanuela Galasso. The World Bank MATCHING TECHNIQUES Technical Track Sessin VI Emanuela Galass The Wrld Bank These slides were develped by Christel Vermeersch and mdified by Emanuela Galass fr the purpse f this wrkshp When can we use

More information

SUPPLEMENTARY MATERIAL GaGa: a simple and flexible hierarchical model for microarray data analysis

SUPPLEMENTARY MATERIAL GaGa: a simple and flexible hierarchical model for microarray data analysis SUPPLEMENTARY MATERIAL GaGa: a simple and flexible hierarchical mdel fr micrarray data analysis David Rssell Department f Bistatistics M.D. Andersn Cancer Center, Hustn, TX 77030, USA rsselldavid@gmail.cm

More information

NOTE ON APPELL POLYNOMIALS

NOTE ON APPELL POLYNOMIALS NOTE ON APPELL POLYNOMIALS I. M. SHEFFER An interesting characterizatin f Appell plynmials by means f a Stieltjes integral has recently been given by Thrne. 1 We prpse t give a secnd such representatin,

More information

4th Indian Institute of Astrophysics - PennState Astrostatistics School July, 2013 Vainu Bappu Observatory, Kavalur. Correlation and Regression

4th Indian Institute of Astrophysics - PennState Astrostatistics School July, 2013 Vainu Bappu Observatory, Kavalur. Correlation and Regression 4th Indian Institute f Astrphysics - PennState Astrstatistics Schl July, 2013 Vainu Bappu Observatry, Kavalur Crrelatin and Regressin Rahul Ry Indian Statistical Institute, Delhi. Crrelatin Cnsider a tw

More information

Hypothesis Tests for One Population Mean

Hypothesis Tests for One Population Mean Hypthesis Tests fr One Ppulatin Mean Chapter 9 Ala Abdelbaki Objective Objective: T estimate the value f ne ppulatin mean Inferential statistics using statistics in rder t estimate parameters We will be

More information

Support-Vector Machines

Support-Vector Machines Supprt-Vectr Machines Intrductin Supprt vectr machine is a linear machine with sme very nice prperties. Haykin chapter 6. See Alpaydin chapter 13 fr similar cntent. Nte: Part f this lecture drew material

More information

ON THE ABSOLUTE CONVERGENCE OF A SERIES ASSOCIATED WITH A FOURIER SERIES. R. MOHANTY and s. mohapatra

ON THE ABSOLUTE CONVERGENCE OF A SERIES ASSOCIATED WITH A FOURIER SERIES. R. MOHANTY and s. mohapatra ON THE ABSOLUTE CONVERGENCE OF A SERIES ASSOCIATED WITH A FOURIER SERIES R. MOHANTY and s. mhapatra 1. Suppse/(i) is integrable L in ( ir, it) peridic with perid 2ir, and that its Furier series at / =

More information

A NOTE ON THE EQUIVAImCE OF SOME TEST CRITERIA. v. P. Bhapkar. University of Horth Carolina. and

A NOTE ON THE EQUIVAImCE OF SOME TEST CRITERIA. v. P. Bhapkar. University of Horth Carolina. and ~ A NOTE ON THE EQUVAmCE OF SOME TEST CRTERA by v. P. Bhapkar University f Hrth Carlina University f Pna nstitute f Statistics Mime Series N. 421 February 1965 This research was supprted by the Mathematics

More information

Inference in the Multiple-Regression

Inference in the Multiple-Regression Sectin 5 Mdel Inference in the Multiple-Regressin Kinds f hypthesis tests in a multiple regressin There are several distinct kinds f hypthesis tests we can run in a multiple regressin. Suppse that amng

More information

, which yields. where z1. and z2

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

More information

Section 6-2: Simplex Method: Maximization with Problem Constraints of the Form ~

Section 6-2: Simplex Method: Maximization with Problem Constraints of the Form ~ Sectin 6-2: Simplex Methd: Maximizatin with Prblem Cnstraints f the Frm ~ Nte: This methd was develped by Gerge B. Dantzig in 1947 while n assignment t the U.S. Department f the Air Frce. Definitin: Standard

More information

(Communicated at the meeting of September 25, 1948.) Izl=6 Izl=6 Izl=6 ~=I. max log lv'i (z)1 = log M;.

(Communicated at the meeting of September 25, 1948.) Izl=6 Izl=6 Izl=6 ~=I. max log lv'i (z)1 = log M;. Mathematics. - On a prblem in the thery f unifrm distributin. By P. ERDÖS and P. TURÁN. 11. Ommunicated by Prf. J. G. VAN DER CORPUT.) Cmmunicated at the meeting f September 5, 1948.) 9. Af ter this first

More information

f t(y)dy f h(x)g(xy) dx fk 4 a. «..

f t(y)dy f h(x)g(xy) dx fk 4 a. «.. CERTAIN OPERATORS AND FOURIER TRANSFORMS ON L2 RICHARD R. GOLDBERG 1. Intrductin. A well knwn therem f Titchmarsh [2] states that if fel2(0, ) and if g is the Furier csine transfrm f/, then G(x)=x~1Jx0g(y)dy

More information

INSTRUMENTAL VARIABLES

INSTRUMENTAL VARIABLES INSTRUMENTAL VARIABLES Technical Track Sessin IV Sergi Urzua University f Maryland Instrumental Variables and IE Tw main uses f IV in impact evaluatin: 1. Crrect fr difference between assignment f treatment

More information

initially lcated away frm the data set never win the cmpetitin, resulting in a nnptimal nal cdebk, [2] [3] [4] and [5]. Khnen's Self Organizing Featur

initially lcated away frm the data set never win the cmpetitin, resulting in a nnptimal nal cdebk, [2] [3] [4] and [5]. Khnen's Self Organizing Featur Cdewrd Distributin fr Frequency Sensitive Cmpetitive Learning with One Dimensinal Input Data Aristides S. Galanpuls and Stanley C. Ahalt Department f Electrical Engineering The Ohi State University Abstract

More information

READING STATECHART DIAGRAMS

READING STATECHART DIAGRAMS READING STATECHART DIAGRAMS Figure 4.48 A Statechart diagram with events The diagram in Figure 4.48 shws all states that the bject plane can be in during the curse f its life. Furthermre, it shws the pssible

More information

A new Type of Fuzzy Functions in Fuzzy Topological Spaces

A new Type of Fuzzy Functions in Fuzzy Topological Spaces IOSR Jurnal f Mathematics (IOSR-JM e-issn: 78-578, p-issn: 39-765X Vlume, Issue 5 Ver I (Sep - Oct06, PP 8-4 wwwisrjurnalsrg A new Type f Fuzzy Functins in Fuzzy Tplgical Spaces Assist Prf Dr Munir Abdul

More information

A New Evaluation Measure. J. Joiner and L. Werner. The problems of evaluation and the needed criteria of evaluation

A New Evaluation Measure. J. Joiner and L. Werner. The problems of evaluation and the needed criteria of evaluation III-l III. A New Evaluatin Measure J. Jiner and L. Werner Abstract The prblems f evaluatin and the needed criteria f evaluatin measures in the SMART system f infrmatin retrieval are reviewed and discussed.

More information

On Huntsberger Type Shrinkage Estimator for the Mean of Normal Distribution ABSTRACT INTRODUCTION

On Huntsberger Type Shrinkage Estimator for the Mean of Normal Distribution ABSTRACT INTRODUCTION Malaysian Jurnal f Mathematical Sciences 4(): 7-4 () On Huntsberger Type Shrinkage Estimatr fr the Mean f Nrmal Distributin Department f Mathematical and Physical Sciences, University f Nizwa, Sultanate

More information

AP Statistics Notes Unit Two: The Normal Distributions

AP Statistics Notes Unit Two: The Normal Distributions AP Statistics Ntes Unit Tw: The Nrmal Distributins Syllabus Objectives: 1.5 The student will summarize distributins f data measuring the psitin using quartiles, percentiles, and standardized scres (z-scres).

More information

Kinetic Model Completeness

Kinetic Model Completeness 5.68J/10.652J Spring 2003 Lecture Ntes Tuesday April 15, 2003 Kinetic Mdel Cmpleteness We say a chemical kinetic mdel is cmplete fr a particular reactin cnditin when it cntains all the species and reactins

More information

MATCHING TECHNIQUES Technical Track Session VI Céline Ferré The World Bank

MATCHING TECHNIQUES Technical Track Session VI Céline Ferré The World Bank MATCHING TECHNIQUES Technical Track Sessin VI Céline Ferré The Wrld Bank When can we use matching? What if the assignment t the treatment is nt dne randmly r based n an eligibility index, but n the basis

More information

Floating Point Method for Solving Transportation. Problems with Additional Constraints

Floating Point Method for Solving Transportation. Problems with Additional Constraints Internatinal Mathematical Frum, Vl. 6, 20, n. 40, 983-992 Flating Pint Methd fr Slving Transprtatin Prblems with Additinal Cnstraints P. Pandian and D. Anuradha Department f Mathematics, Schl f Advanced

More information

Chapter 3: Cluster Analysis

Chapter 3: Cluster Analysis Chapter 3: Cluster Analysis } 3.1 Basic Cncepts f Clustering 3.1.1 Cluster Analysis 3.1. Clustering Categries } 3. Partitining Methds 3..1 The principle 3.. K-Means Methd 3..3 K-Medids Methd 3..4 CLARA

More information

UNIV1"'RSITY OF NORTH CAROLINA Department of Statistics Chapel Hill, N. C. CUMULATIVE SUM CONTROL CHARTS FOR THE FOLDED NORMAL DISTRIBUTION

UNIV1'RSITY OF NORTH CAROLINA Department of Statistics Chapel Hill, N. C. CUMULATIVE SUM CONTROL CHARTS FOR THE FOLDED NORMAL DISTRIBUTION UNIV1"'RSITY OF NORTH CAROLINA Department f Statistics Chapel Hill, N. C. CUMULATIVE SUM CONTROL CHARTS FOR THE FOLDED NORMAL DISTRIBUTION by N. L. Jlmsn December 1962 Grant N. AFOSR -62..148 Methds f

More information

MAKING DOUGHNUTS OF COHEN REALS

MAKING DOUGHNUTS OF COHEN REALS MAKING DUGHNUTS F CHEN REALS Lrenz Halbeisen Department f Pure Mathematics Queen s University Belfast Belfast BT7 1NN, Nrthern Ireland Email: halbeis@qub.ac.uk Abstract Fr a b ω with b \ a infinite, the

More information

Chapter 2 GAUSS LAW Recommended Problems:

Chapter 2 GAUSS LAW Recommended Problems: Chapter GAUSS LAW Recmmended Prblems: 1,4,5,6,7,9,11,13,15,18,19,1,7,9,31,35,37,39,41,43,45,47,49,51,55,57,61,6,69. LCTRIC FLUX lectric flux is a measure f the number f electric filed lines penetrating

More information

Dead-beat controller design

Dead-beat controller design J. Hetthéssy, A. Barta, R. Bars: Dead beat cntrller design Nvember, 4 Dead-beat cntrller design In sampled data cntrl systems the cntrller is realised by an intelligent device, typically by a PLC (Prgrammable

More information

Determining the Accuracy of Modal Parameter Estimation Methods

Determining the Accuracy of Modal Parameter Estimation Methods Determining the Accuracy f Mdal Parameter Estimatin Methds by Michael Lee Ph.D., P.E. & Mar Richardsn Ph.D. Structural Measurement Systems Milpitas, CA Abstract The mst cmmn type f mdal testing system

More information

SIZE BIAS IN LINE TRANSECT SAMPLING: A FIELD TEST. Mark C. Otto Statistics Research Division, Bureau of the Census Washington, D.C , U.S.A.

SIZE BIAS IN LINE TRANSECT SAMPLING: A FIELD TEST. Mark C. Otto Statistics Research Division, Bureau of the Census Washington, D.C , U.S.A. SIZE BIAS IN LINE TRANSECT SAMPLING: A FIELD TEST Mark C. Ott Statistics Research Divisin, Bureau f the Census Washingtn, D.C. 20233, U.S.A. and Kenneth H. Pllck Department f Statistics, Nrth Carlina State

More information

Tree Structured Classifier

Tree Structured Classifier Tree Structured Classifier Reference: Classificatin and Regressin Trees by L. Breiman, J. H. Friedman, R. A. Olshen, and C. J. Stne, Chapman & Hall, 98. A Medical Eample (CART): Predict high risk patients

More information

Lecture 2: Supervised vs. unsupervised learning, bias-variance tradeoff

Lecture 2: Supervised vs. unsupervised learning, bias-variance tradeoff Lecture 2: Supervised vs. unsupervised learning, bias-variance tradeff Reading: Chapter 2 STATS 202: Data mining and analysis September 27, 2017 1 / 20 Supervised vs. unsupervised learning In unsupervised

More information

Lecture 2: Supervised vs. unsupervised learning, bias-variance tradeoff

Lecture 2: Supervised vs. unsupervised learning, bias-variance tradeoff Lecture 2: Supervised vs. unsupervised learning, bias-variance tradeff Reading: Chapter 2 STATS 202: Data mining and analysis September 27, 2017 1 / 20 Supervised vs. unsupervised learning In unsupervised

More information

CAUSAL INFERENCE. Technical Track Session I. Phillippe Leite. The World Bank

CAUSAL INFERENCE. Technical Track Session I. Phillippe Leite. The World Bank CAUSAL INFERENCE Technical Track Sessin I Phillippe Leite The Wrld Bank These slides were develped by Christel Vermeersch and mdified by Phillippe Leite fr the purpse f this wrkshp Plicy questins are causal

More information

Perfrmance f Sensitizing Rules n Shewhart Cntrl Charts with Autcrrelated Data Key Wrds: Autregressive, Mving Average, Runs Tests, Shewhart Cntrl Chart

Perfrmance f Sensitizing Rules n Shewhart Cntrl Charts with Autcrrelated Data Key Wrds: Autregressive, Mving Average, Runs Tests, Shewhart Cntrl Chart Perfrmance f Sensitizing Rules n Shewhart Cntrl Charts with Autcrrelated Data Sandy D. Balkin Dennis K. J. Lin y Pennsylvania State University, University Park, PA 16802 Sandy Balkin is a graduate student

More information

Differentiation Applications 1: Related Rates

Differentiation Applications 1: Related Rates Differentiatin Applicatins 1: Related Rates 151 Differentiatin Applicatins 1: Related Rates Mdel 1: Sliding Ladder 10 ladder y 10 ladder 10 ladder A 10 ft ladder is leaning against a wall when the bttm

More information

Pure adaptive search for finite global optimization*

Pure adaptive search for finite global optimization* Mathematical Prgramming 69 (1995) 443-448 Pure adaptive search fr finite glbal ptimizatin* Z.B. Zabinskya.*, G.R. Wd b, M.A. Steel c, W.P. Baritmpa c a Industrial Engineering Prgram, FU-20. University

More information

AN INTERMITTENTLY USED SYSTEM WITH PREVENTIVE MAINTENANCE

AN INTERMITTENTLY USED SYSTEM WITH PREVENTIVE MAINTENANCE J. Operatins Research Sc. f Japan V!. 15, N. 2, June 1972. 1972 The Operatins Research Sciety f Japan AN INTERMITTENTLY USED SYSTEM WITH PREVENTIVE MAINTENANCE SHUNJI OSAKI University f Suthern Califrnia

More information

What is Statistical Learning?

What is Statistical Learning? What is Statistical Learning? Sales 5 10 15 20 25 Sales 5 10 15 20 25 Sales 5 10 15 20 25 0 50 100 200 300 TV 0 10 20 30 40 50 Radi 0 20 40 60 80 100 Newspaper Shwn are Sales vs TV, Radi and Newspaper,

More information

A solution of certain Diophantine problems

A solution of certain Diophantine problems A slutin f certain Diphantine prblems Authr L. Euler* E7 Nvi Cmmentarii academiae scientiarum Petrplitanae 0, 1776, pp. 8-58 Opera Omnia: Series 1, Vlume 3, pp. 05-17 Reprinted in Cmmentat. arithm. 1,

More information

CS 477/677 Analysis of Algorithms Fall 2007 Dr. George Bebis Course Project Due Date: 11/29/2007

CS 477/677 Analysis of Algorithms Fall 2007 Dr. George Bebis Course Project Due Date: 11/29/2007 CS 477/677 Analysis f Algrithms Fall 2007 Dr. Gerge Bebis Curse Prject Due Date: 11/29/2007 Part1: Cmparisn f Srting Algrithms (70% f the prject grade) The bjective f the first part f the assignment is

More information

1996 Engineering Systems Design and Analysis Conference, Montpellier, France, July 1-4, 1996, Vol. 7, pp

1996 Engineering Systems Design and Analysis Conference, Montpellier, France, July 1-4, 1996, Vol. 7, pp THE POWER AND LIMIT OF NEURAL NETWORKS T. Y. Lin Department f Mathematics and Cmputer Science San Jse State University San Jse, Califrnia 959-003 tylin@cs.ssu.edu and Bereley Initiative in Sft Cmputing*

More information

Lecture 13: Markov Chain Monte Carlo. Gibbs sampling

Lecture 13: Markov Chain Monte Carlo. Gibbs sampling Lecture 13: Markv hain Mnte arl Gibbs sampling Gibbs sampling Markv chains 1 Recall: Apprximate inference using samples Main idea: we generate samples frm ur Bayes net, then cmpute prbabilities using (weighted)

More information

2004 AP CHEMISTRY FREE-RESPONSE QUESTIONS

2004 AP CHEMISTRY FREE-RESPONSE QUESTIONS 2004 AP CHEMISTRY FREE-RESPONSE QUESTIONS 6. An electrchemical cell is cnstructed with an pen switch, as shwn in the diagram abve. A strip f Sn and a strip f an unknwn metal, X, are used as electrdes.

More information

Thermodynamics Partial Outline of Topics

Thermodynamics Partial Outline of Topics Thermdynamics Partial Outline f Tpics I. The secnd law f thermdynamics addresses the issue f spntaneity and invlves a functin called entrpy (S): If a prcess is spntaneus, then Suniverse > 0 (2 nd Law!)

More information

Review Problems 3. Four FIR Filter Types

Review Problems 3. Four FIR Filter Types Review Prblems 3 Fur FIR Filter Types Fur types f FIR linear phase digital filters have cefficients h(n fr 0 n M. They are defined as fllws: Type I: h(n = h(m-n and M even. Type II: h(n = h(m-n and M dd.

More information

MATHEMATICS SYLLABUS SECONDARY 5th YEAR

MATHEMATICS SYLLABUS SECONDARY 5th YEAR Eurpean Schls Office f the Secretary-General Pedaggical Develpment Unit Ref. : 011-01-D-8-en- Orig. : EN MATHEMATICS SYLLABUS SECONDARY 5th YEAR 6 perid/week curse APPROVED BY THE JOINT TEACHING COMMITTEE

More information

FINITE BOOLEAN ALGEBRA. 1. Deconstructing Boolean algebras with atoms. Let B = <B,,,,,0,1> be a Boolean algebra and c B.

FINITE BOOLEAN ALGEBRA. 1. Deconstructing Boolean algebras with atoms. Let B = <B,,,,,0,1> be a Boolean algebra and c B. FINITE BOOLEAN ALGEBRA 1. Decnstructing Blean algebras with atms. Let B = be a Blean algebra and c B. The ideal generated by c, (c], is: (c] = {b B: b c} The filter generated by c, [c), is:

More information

Chemistry 20 Lesson 11 Electronegativity, Polarity and Shapes

Chemistry 20 Lesson 11 Electronegativity, Polarity and Shapes Chemistry 20 Lessn 11 Electrnegativity, Plarity and Shapes In ur previus wrk we learned why atms frm cvalent bnds and hw t draw the resulting rganizatin f atms. In this lessn we will learn (a) hw the cmbinatin

More information

NOTE ON THE ANALYSIS OF A RANDOMIZED BLOCK DESIGN. Junjiro Ogawa University of North Carolina

NOTE ON THE ANALYSIS OF A RANDOMIZED BLOCK DESIGN. Junjiro Ogawa University of North Carolina NOTE ON THE ANALYSIS OF A RANDOMIZED BLOCK DESIGN by Junjir Ogawa University f Nrth Carlina This research was supprted by the Office f Naval Research under Cntract N. Nnr-855(06) fr research in prbability

More information

Source Coding and Compression

Source Coding and Compression Surce Cding and Cmpressin Heik Schwarz Cntact: Dr.-Ing. Heik Schwarz heik.schwarz@hhi.fraunhfer.de Heik Schwarz Surce Cding and Cmpressin September 22, 2013 1 / 60 PartI: Surce Cding Fundamentals Heik

More information

Simple Linear Regression (single variable)

Simple Linear Regression (single variable) Simple Linear Regressin (single variable) Intrductin t Machine Learning Marek Petrik January 31, 2017 Sme f the figures in this presentatin are taken frm An Intrductin t Statistical Learning, with applicatins

More information

The blessing of dimensionality for kernel methods

The blessing of dimensionality for kernel methods fr kernel methds Building classifiers in high dimensinal space Pierre Dupnt Pierre.Dupnt@ucluvain.be Classifiers define decisin surfaces in sme feature space where the data is either initially represented

More information

Pressure And Entropy Variations Across The Weak Shock Wave Due To Viscosity Effects

Pressure And Entropy Variations Across The Weak Shock Wave Due To Viscosity Effects Pressure And Entrpy Variatins Acrss The Weak Shck Wave Due T Viscsity Effects OSTAFA A. A. AHOUD Department f athematics Faculty f Science Benha University 13518 Benha EGYPT Abstract:-The nnlinear differential

More information

Introduction: A Generalized approach for computing the trajectories associated with the Newtonian N Body Problem

Introduction: A Generalized approach for computing the trajectories associated with the Newtonian N Body Problem A Generalized apprach fr cmputing the trajectries assciated with the Newtnian N Bdy Prblem AbuBar Mehmd, Syed Umer Abbas Shah and Ghulam Shabbir Faculty f Engineering Sciences, GIK Institute f Engineering

More information

MODULE FOUR. This module addresses functions. SC Academic Elementary Algebra Standards:

MODULE FOUR. This module addresses functions. SC Academic Elementary Algebra Standards: MODULE FOUR This mdule addresses functins SC Academic Standards: EA-3.1 Classify a relatinship as being either a functin r nt a functin when given data as a table, set f rdered pairs, r graph. EA-3.2 Use

More information

Localized Model Selection for Regression

Localized Model Selection for Regression Lcalized Mdel Selectin fr Regressin Yuhng Yang Schl f Statistics University f Minnesta Church Street S.E. Minneaplis, MN 5555 May 7, 007 Abstract Research n mdel/prcedure selectin has fcused n selecting

More information

ON-LINE PROCEDURE FOR TERMINATING AN ACCELERATED DEGRADATION TEST

ON-LINE PROCEDURE FOR TERMINATING AN ACCELERATED DEGRADATION TEST Statistica Sinica 8(1998), 207-220 ON-LINE PROCEDURE FOR TERMINATING AN ACCELERATED DEGRADATION TEST Hng-Fwu Yu and Sheng-Tsaing Tseng Natinal Taiwan University f Science and Technlgy and Natinal Tsing-Hua

More information

Marginal Conceptual Predictive Statistic for Mixed Model Selection

Marginal Conceptual Predictive Statistic for Mixed Model Selection Open Jurnal f Statistics, 06, 6, 39-53 Published Online April 06 in Sci http://wwwscirprg/jurnal/js http://dxdirg/0436/js0660 Marginal Cnceptual Predictive Statistic fr Mixed Mdel Selectin Cheng Wenren,

More information

45 K. M. Dyaknv Garsia nrm kfk G = sup zd jfj d z ;jf(z)j = dened riginally fr f L (T m), is in fact an equivalent nrm n BMO. We shall als be cncerned

45 K. M. Dyaknv Garsia nrm kfk G = sup zd jfj d z ;jf(z)j = dened riginally fr f L (T m), is in fact an equivalent nrm n BMO. We shall als be cncerned Revista Matematica Iberamericana Vl. 5, N. 3, 999 Abslute values f BMOA functins Knstantin M. Dyaknv Abstract. The paper cntains a cmplete characterizatin f the mduli f BMOA functins. These are described

More information

COMP 551 Applied Machine Learning Lecture 11: Support Vector Machines

COMP 551 Applied Machine Learning Lecture 11: Support Vector Machines COMP 551 Applied Machine Learning Lecture 11: Supprt Vectr Machines Instructr: (jpineau@cs.mcgill.ca) Class web page: www.cs.mcgill.ca/~jpineau/cmp551 Unless therwise nted, all material psted fr this curse

More information

Equilibrium of Stress

Equilibrium of Stress Equilibrium f Stress Cnsider tw perpendicular planes passing thrugh a pint p. The stress cmpnents acting n these planes are as shwn in ig. 3.4.1a. These stresses are usuall shwn tgether acting n a small

More information

Math Foundations 20 Work Plan

Math Foundations 20 Work Plan Math Fundatins 20 Wrk Plan Units / Tpics 20.8 Demnstrate understanding f systems f linear inequalities in tw variables. Time Frame December 1-3 weeks 6-10 Majr Learning Indicatrs Identify situatins relevant

More information

Pattern Recognition 2014 Support Vector Machines

Pattern Recognition 2014 Support Vector Machines Pattern Recgnitin 2014 Supprt Vectr Machines Ad Feelders Universiteit Utrecht Ad Feelders ( Universiteit Utrecht ) Pattern Recgnitin 1 / 55 Overview 1 Separable Case 2 Kernel Functins 3 Allwing Errrs (Sft

More information

The Equation αsin x+ βcos family of Heron Cyclic Quadrilaterals

The Equation αsin x+ βcos family of Heron Cyclic Quadrilaterals The Equatin sin x+ βcs x= γ and a family f Hern Cyclic Quadrilaterals Knstantine Zelatr Department Of Mathematics Cllege Of Arts And Sciences Mail Stp 94 University Of Tled Tled,OH 43606-3390 U.S.A. Intrductin

More information

Measurement Errors in Quantile Regression Models

Measurement Errors in Quantile Regression Models Measurement Errrs in Quantile Regressin Mdels Sergi Firp Antni F. Galva Suyng Sng June 30, 2015 Abstract This paper develps estimatin and inference fr quantile regressin mdels with measurement errrs. We

More information

A Simple Set of Test Matrices for Eigenvalue Programs*

A Simple Set of Test Matrices for Eigenvalue Programs* Simple Set f Test Matrices fr Eigenvalue Prgrams* By C. W. Gear** bstract. Sets f simple matrices f rder N are given, tgether with all f their eigenvalues and right eigenvectrs, and simple rules fr generating

More information

THE DEVELOPMENT OF AND GAPS IN THE THEORY OF PRODUCTS OF INITIALLY M-COMPACT SPACES

THE DEVELOPMENT OF AND GAPS IN THE THEORY OF PRODUCTS OF INITIALLY M-COMPACT SPACES Vlume 6, 1981 Pages 99 113 http://tplgy.auburn.edu/tp/ THE DEVELOPMENT OF AND GAPS IN THE THEORY OF PRODUCTS OF INITIALLY M-COMPACT SPACES by R. M. Stephensn, Jr. Tplgy Prceedings Web: http://tplgy.auburn.edu/tp/

More information

Revision: August 19, E Main Suite D Pullman, WA (509) Voice and Fax

Revision: August 19, E Main Suite D Pullman, WA (509) Voice and Fax .7.4: Direct frequency dmain circuit analysis Revisin: August 9, 00 5 E Main Suite D Pullman, WA 9963 (509) 334 6306 ice and Fax Overview n chapter.7., we determined the steadystate respnse f electrical

More information

Comparing Several Means: ANOVA. Group Means and Grand Mean

Comparing Several Means: ANOVA. Group Means and Grand Mean STAT 511 ANOVA and Regressin 1 Cmparing Several Means: ANOVA Slide 1 Blue Lake snap beans were grwn in 12 pen-tp chambers which are subject t 4 treatments 3 each with O 3 and SO 2 present/absent. The ttal

More information

Sections 15.1 to 15.12, 16.1 and 16.2 of the textbook (Robbins-Miller) cover the materials required for this topic.

Sections 15.1 to 15.12, 16.1 and 16.2 of the textbook (Robbins-Miller) cover the materials required for this topic. Tpic : AC Fundamentals, Sinusidal Wavefrm, and Phasrs Sectins 5. t 5., 6. and 6. f the textbk (Rbbins-Miller) cver the materials required fr this tpic.. Wavefrms in electrical systems are current r vltage

More information

PSU GISPOPSCI June 2011 Ordinary Least Squares & Spatial Linear Regression in GeoDa

PSU GISPOPSCI June 2011 Ordinary Least Squares & Spatial Linear Regression in GeoDa There are tw parts t this lab. The first is intended t demnstrate hw t request and interpret the spatial diagnstics f a standard OLS regressin mdel using GeDa. The diagnstics prvide infrmatin abut the

More information

CHAPTER 3 INEQUALITIES. Copyright -The Institute of Chartered Accountants of India

CHAPTER 3 INEQUALITIES. Copyright -The Institute of Chartered Accountants of India CHAPTER 3 INEQUALITIES Cpyright -The Institute f Chartered Accuntants f India INEQUALITIES LEARNING OBJECTIVES One f the widely used decisin making prblems, nwadays, is t decide n the ptimal mix f scarce

More information

B. Definition of an exponential

B. Definition of an exponential Expnents and Lgarithms Chapter IV - Expnents and Lgarithms A. Intrductin Starting with additin and defining the ntatins fr subtractin, multiplicatin and divisin, we discvered negative numbers and fractins.

More information

[COLLEGE ALGEBRA EXAM I REVIEW TOPICS] ( u s e t h i s t o m a k e s u r e y o u a r e r e a d y )

[COLLEGE ALGEBRA EXAM I REVIEW TOPICS] ( u s e t h i s t o m a k e s u r e y o u a r e r e a d y ) (Abut the final) [COLLEGE ALGEBRA EXAM I REVIEW TOPICS] ( u s e t h i s t m a k e s u r e y u a r e r e a d y ) The department writes the final exam s I dn't really knw what's n it and I can't very well

More information

Eric Klein and Ning Sa

Eric Klein and Ning Sa Week 12. Statistical Appraches t Netwrks: p1 and p* Wasserman and Faust Chapter 15: Statistical Analysis f Single Relatinal Netwrks There are fur tasks in psitinal analysis: 1) Define Equivalence 2) Measure

More information

On Out-of-Sample Statistics for Financial Time-Series

On Out-of-Sample Statistics for Financial Time-Series On Out-f-Sample Statistics fr Financial Time-Series Françis Gingras Yshua Bengi Claude Nadeau CRM-2585 January 1999 Département de physique, Université de Mntréal Labratire d infrmatique des systèmes adaptatifs,

More information

Least Squares Optimal Filtering with Multirate Observations

Least Squares Optimal Filtering with Multirate Observations Prc. 36th Asilmar Cnf. n Signals, Systems, and Cmputers, Pacific Grve, CA, Nvember 2002 Least Squares Optimal Filtering with Multirate Observatins Charles W. herrien and Anthny H. Hawes Department f Electrical

More information

ENSC Discrete Time Systems. Project Outline. Semester

ENSC Discrete Time Systems. Project Outline. Semester ENSC 49 - iscrete Time Systems Prject Outline Semester 006-1. Objectives The gal f the prject is t design a channel fading simulatr. Upn successful cmpletin f the prject, yu will reinfrce yur understanding

More information

UG Course Outline EC2203: Quantitative Methods II 2017/18

UG Course Outline EC2203: Quantitative Methods II 2017/18 UG Curse Outline EC2203: Quantitative Methds II 2017/18 Autumn: Instructr: Pierre0-Olivier Frtin Office: Hrtn H214 Phne: +44 (0) 1784 276474 E-mail: pierre-livier.frtin@rhul.ac.uk Office hurs: Tuesdays

More information

Chapter 9 Vector Differential Calculus, Grad, Div, Curl

Chapter 9 Vector Differential Calculus, Grad, Div, Curl Chapter 9 Vectr Differential Calculus, Grad, Div, Curl 9.1 Vectrs in 2-Space and 3-Space 9.2 Inner Prduct (Dt Prduct) 9.3 Vectr Prduct (Crss Prduct, Outer Prduct) 9.4 Vectr and Scalar Functins and Fields

More information

Probability, Random Variables, and Processes. Probability

Probability, Random Variables, and Processes. Probability Prbability, Randm Variables, and Prcesses Prbability Prbability Prbability thery: branch f mathematics fr descriptin and mdelling f randm events Mdern prbability thery - the aximatic definitin f prbability

More information

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

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

More information