epub WU Institutional Repository

Size: px
Start display at page:

Download "epub WU Institutional Repository"

Transcription

1 epub WU Institutional Repository Michael A. Hauser Maximum Likelihoo Estimators for ARMA an ARFIMA Moels. A Monte Carlo Stuy. Paper Original Citation: Hauser, Michael A. (998) Maximum Likelihoo Estimators for ARMA an ARFIMA Moels. A Monte Carlo Stuy. Preprint Series / Department of Applie Statistics an Data Processing, 22. Department of Statistics an Mathematics, Abt. f. Angewante Statistik u. Datenverarbeitung, WU Vienna University of Economics an Business, Vienna. This version is available at: Available in epub WU : July 26 epub WU, the institutional repository of the WU Vienna University of Economics an Business, is provie by the University Library an the IT-Services. The aim is to enable open access to the scholarly output of the WU.

2 Maximum Likelihoo Estimators for ARMA an ARFIMA Moels: A Monte Carlo Stuy Michael Hauser Department of Applie Statistics an Data Processing Wirtschaftsuniversität Wien Preprint Series Preprint 22 May 998

3 Maximum Likelihoo Estimators for ARMA an ARFIMA Moels: A Monte Carlo Stuy Michael A. Hauser University of Economics an Business Aministration, Department of Statistics, Vienna First raft December 997. This version May 998. Forthcoming Journal of Statistical Planning an Inference Abstract We analyze by simulation the properties of two time omain an two frequency omain estimators for low orer autoregressive fractionally integrate moving average Gaussian moels, ARFIMA (p; ; q). The estimators consiere are the exact maximum likelihoo for emeane ata,, the associate moie prole likelihoo,, an the Whittle estimator with,, an without tapere ata,. Length of the series is. The estimators are compare in terms of pile-up eect, mean square error, bias, an empirical conence level. The tapere version of the Whittle likelihoo turns out to be a reliable estimator for ARMA an ARFIMA moels. Its small losses in performance in case of \well-behave" moels are compensate suciently in more \icult" moels. The moie prole likelihoo is an alternative to the but is computationally more emaning. It is either equivalent to the or more favorable than the. For fractionally integrate moels, particularly, it ominates clearly the. The has serious eciencies for large ranges of parameters, an so cannot be recommene in general. The, on the other han, shoul only be use with care for fractionally integrate moels ue to its potential large negative bias of the fractional integration parameter. In general, one shoul procee with caution for ARMA(,) moels with almost canceling roots, an, in particular, in case of the an the for inference in the vicinity of a moving average root of +. Keywors: Fractional integration, Whittle likelihoo, moie prole likelihoo, ata taper, pile-up eect Acknowlegments: I want to express my thanks to Peter Bloomel, Sastry Pantula, Beneikt Potscher, an iscussants at the International Workshop on Long Range Depenence, Brisbane, January 997, an two anonymous referees. Short title: Maximum Likelihoo Estimators for ARFIMA Moels Aress of the author: Department of Statistics, University of Economics an Business Aministration, Augasse 2-6, A-9 Vienna, Austria. Tel.: /4759, FAX: /738, Michael.Hauser@wu-wien.ac.at

4 Introuction In this paper we analyze, by means of simulation, the properties of four versions of maximum likelihoo estimators for tting autoregressive fractionally integrate moving average, ARFIMA, time series moels. The comparison of the small sample properties is of particular interest since ierent estimators may lea to substantial istinct conclusions. E.g. if the Whittle estimate or the moie prole likelihoo is use for tting a class of low orer ARFIMA moels to the frequently investigate postwar quarterly U.S. GNP growth rates Akaike's information criterion will choose a non fractional moel. On the other han, if the same proceure is unertaken with the exact Gaussian likelihoo, it results in a fractionally integrate moel with a highly signicant fractional integration parameter. See Sowell(992b) an Hauser(995). Thus, espite the fact that the estimators are asymptotically equivalent the question remains which estimator is more reliable in small samples. We assume that the series y t is a sample of length n of a weakly stationary process (in iscrete time) which amits a Wol representation y t = + (L)u t = + ( + L + 2 L )u t : where L enotes the backwar shift operator. The impulse response coecients satisfy P = 2 <, an the u t are Gaussian white noise with variance 2 u. An, y t obeys to an ARFIMA(p; ; q) moel with 2 [ 2 ; 2 ). (L)( L) (y t ) = (L)u t ; where (L) = L p L p an (L) = L q L q ; are polynomials in the backwar shift operator an is the fractional integration parameter. The roots of the polynomials (z) an (z) are assume to lie outsie the unit circle. Then the process (y t ) is stationary for < 2 an invertible for > (Bloomel, 985). The spectrum f() = 2 u j(exp( i))j 2 2 j(exp( i))j 2 j exp( i)j 2 ; is innite at the origin for > an zero for <. Long memory is associate with >. If <, the process is sai to have intermeiate memory.

5 Two generic maximum likelihoo proceures for stationary Gaussian series for the parameter vector = ( ; ; p ; ; ; ; q ) are available: the (approximative) spectral (Whittle) maximum likelihoo, an the exact Gaussian maximum likelihoo metho. Both methos yiel p n-consistent, asymptotically normal an asymptotically ecient parameter estimates. See Fox an Taqqu(986), Dahlhaus(989) an Giraitis an Surgailis(99) for the Whittle estimator, an Dahlhaus(989) an An, Bloomel an Pantula(992) for the exact maximum likelihoo. The small sample properties of four estimators will be investigate: the exact Gaussian likelihoo, the moie prole likelihoo erive by An an Bloomel(993) accoring to the proposal of Cox an Rei(987), the Whittle likelihoo an two versions thereof using tapere ata. Having economic applications in min we restrict our stuy to low orer ARMA an ARFIMA moels. The criteria for comparison are the mean square error, bias, an the empirical conence level of the 95% conence interval. In aition, the estimators will be investigate with respect of the occurrence of the pile-up eect, which has been intensively iscusse in the literature for the exact maximum likelihoo metho. Our stuy is a signicant extension to the former simulation experiments performe by Ansley an Newbol(98), Boes et al.(989) an Cheung an Diebol(994). No Monte Carlo stuy seems to be available which compares the small sample properties of maximum likelihoo estimators even for ARMA moels. An, estimators using tapere ata have been applie almost exclusively to AR(p) or FI() moels. Section 2 enes the estimators an summarizes their small sample properties as far as they are known an relevant for our stuy. Section 3 summarizes the Monte Carlo results. Computational aspects are iscusse in Section 3.. For the Monte Carlo results of the pile-up eect see Section 3.2. The ARFIMA moels chosen an the criteria for the comparison of the estimators are given in Section 3.3. The properties of the four estimators are presente in Section 3.4. Section 3.5 gives a heuristic explanation of the systematic negative bias of the exact maximum likelihoo estimates of. Section 4 conclues. 2

6 2 The estimators The exact maximum likelihoo proceure, The exact Gaussian log-likelihoo is given by log L E (; ; 2 u) = 2 [n log(2) + log et() + (Y l) (Y l)] () where = (; 2 u) is the n n covariance matrix epening on an 2 u. n enotes the sample size, Y = (y ;... ; y n ) an l = (;... ; ). We will investigate the reuce prole likelihoo where is replace by the sample mean y, an where the solution of the maximization with respect to 2 u is obtaine analytically. The reuce (or concentrate) likelihoo is then with 2 u = 2 u log L E() = [n log(2) + n log(2 u ) + log et( ) 2 ~ + n] (2) an 2 u = n (Y yl) ~ (Y yl). ~ is ene by the relation = 2 u ~, so that ~ = ~() is a function of alone. The resulting estimator is enote by. Two small sample properties of the may be of importance: The pile-up eect arises when the unerlying MA moel has a root \close" to the unit circle. Then the estimates yiel roots on the unit circle with a positive probability. (Cp. Cryer an Leolter(98) an Anerson an Takemura(986).) On the other han, it is known that the estimate tens to give a negatively biase parameter for pure fractionally integrate processes (Li an McLeo, 986). This is not in eect if the true mean is known an is use for the mean correction of the ata (Cheung an Diebol, 994). The moie prole likelihoo, The moie prole likelihoo is base on the iea to correct the parameter estimates of interest (^a) for secon orer eects ue to nuisance parameters (b) in the moel. (Cp. Cox an Rei(987).) Thereby a transformation is sought which makes b orthogonal to a. For ARFIMA moels the nuisance parameters are chosen as b = (; y), 2 an the parameters of interest as a = = ( ; ; p ; ; ; ; q ). An an Bloomel(993) give the solution for the moie prole likelihoo base on the Gaussian exact maximum likelihoo. The orthogonal vector results as = (; ) with = 2 y (et R) =n, where R = = 2 y is the 3

7 correlation matrix. Remarkably, turns out to be orthogonal to. The without constants is then (in a notation similar to An an Bloomel) log L M(; ^ 2 u; ^) = ( n 2 ) log et (R) 2 log (l R l) + ( + 2 n 2 )(Y ^l) R (Y ^l)] (3) Our actual implementation replaces ^ by the sample mean y, an ^ 2 y by the maximum likelihoo estimator ^ 2 y = (Y n yl) R (Y yl). ^ 2 u results from the relation 2 y = 2 P 2 u i i. An an Bloomel also oer a small Monte Carlo stuy illustrating for some selecte low orer ARFIMA moels that the successfully eliminates the bias in the estimates. The Whittle likelihoo, We ene the Whittle likelihoo as mx log L W (; u) 2 = j= log f( j j ; 2 u) 2 mx j= I( j ) f( j j ; 2 u) (4) where I( j ) enotes the perioogram at the j-th Fourier frequency, j ( j = 2 j n, j = ;... ; m) I( j ) = n j nx m is the largest integer containe in (n )=2. y t exp( it j )j 2 (5) t= It is the iscrete time version of the Whittle function (cp. Dahlhaus(988) an Robinson(99)). In the ARMA case it may be interprete e.g. as the likelihoo associate with the asymptotic istribution of the perioogram (cp. Brockwell an Davis, 99, p.347f). Other interpretations are given in Dahlhaus(988) an Parzen(983, Sec. 3). On the other han, if the term P m j= log f( j j ; 2 u) is roppe the asymptotic properties remain the same (Fox an Taqqu, 986), an it becomes the Yule-Walker estimator for AR(p) processes (Dzhaparize, 986, p.6f). The reuce form of L W with respect to the error variance 2 u is with 2 u = 2 u log L W () = m log(2) m log[ I( j ) m j= 2 u mx = m j= 4 mx g( j ) ] mx I( j ) g( j ) log g( j ) m (6) j= (7)

8 where f() = 2 u g()=(2) with g() = g( j ). This estimator is enote by. The pile-up eect has not been consiere for the Whittle estimate before. In the line with Anerson an Takemura(986) we prove that the likelihoo of a MA() process exhibits a local maximum at =. The same hols contrary to the or for AR() processes at =. The local maximum may turn out as a global one in nite samples, so that parameter estimates of are obtaine with a positive probability. These probabilities are etermine by Monte Carlo simulations in Sec Actually, we only investigate the rst orer conitions. They are for a moel with a single parameter ( may be or ) (@ log L W )=(@) =, or more explicitly X I X j u 2 j g 2 j g = with g j = g( j ) an I( j ) = I j. After some manipulations we obtain X I j j=@ X ] = g j j j g j I j g j [ X k =@ g k ]: So, if the terms in brackets are constant an equal, the rst orer conition hols. In case of MA() processes g() = 2 cos() + 2, = 2[ cos()], g = 8 if =. Both bracket terms are for = either +m or m, an may be cancele. The rst orer conition hols for =. In case of AR() processes g() = =[ 2 cos()+ 2 = ( 2)[ cos()] = [ 2 cos() + 2 ] 2, an g = 8 if =. An, the same hols as above. It is known that Yule-Walker estimates are rather ba for short series, an if the roots of the corresponing characteristic equation are close to the unit circle (Priestley, 98, p.35). So, one may expect similar properties to hol for the Whittle estimates. On the other han, Hauser(995) reports an avantage of the with respect to the for pure fractionally integrate moels an some higher orer ARFIMA moels in small samples. The gives essentially unbiase estimates an smaller mean square errors for in most cases. 5

9 The Whittle likelihoo for tapere ata, Dahlhaus(988) shows that tapering reuces the leakage eect of the perioogram as estimate of the true spectrum. He ns that the new estimate competes well with the Burg estimate for an AR(4) moel where roots of the characteristic equation are complex an close to the unit circle. A tapere series is ene by y T t = h t y t where h t is the ata taper, in our case the Tukey-Hanning taper h t = 8 >< >: 2 [ cosf(t 2 )=lg] t = ;... ; l t = l + ;... ; n l : (8) 2 [ cosf(n t + 2 )=lg] t = n l + ;... ; n The proportion of the ata which is altere by this taper is 2 = 2 l=n. We choose the variable taper of Zhang(99) who proposes to use = 2= p n. The tapere time series is then use to construct moie perioogram orinates, I T ( j ), I T ( j ) = H 2 n j nx h t (y t y) exp( i t j )j 2 (9) t= with H 2 = P t h 2 t. Replacing I( j ) in the function by I T ( j ) yiels the Whittle likelihoo for tapere ata, enote as. There are several stuies concerne with the use of ata tapers ealing essentially only with AR moels: Pukkila an Nyquist(985), Kang(987), Hurvich(988), or Zhang(99), among others. The general conclusions are that the type of taper is not of much importance, nor oes the amount of tapering aect the results consierably. However, some tapering shoul be performe. Cheung an Diebol(994), on the other han, investigate the small sample properties of the approximative Whittle estimate (without the log f-term) for pure FI() moels. They n it to be slightly inferior to the regarless whether tapere or non tapere ata are use. 3 Small sample behavior of the estimators 3. Computational aspects All simulate series throughout the paper are generate via the Durbin-Levinson algorithm. Thereby the true autocovariances (cp. Sowell, 992a) an Gaussian innovations 6

10 with unit variance (Hoermann an Deringer, 99) are use. This metho is exact. For the generation of the uniformly istribute input variates the TT8, a twiste GFSR generator (Matsumoto an Kurita, 994) is use. It has a perio of 2 8 an excellent equiistribution properties up to imension 25. Due to numerical problems of the an when the parameters approach the non stationarity region we have to restrict the estimate AR an parameters. We set max j=z AR j < :9965, where z AR is the root of the characteristic equation for the AR polynomial. The maximal is The program use for calculating the estimates of the, an so also of the, is base on the FORTRAN coe supplie by F. Sowell. The spectral estimates, on the other han, are not boune away from the unit circle. If estimate roots lie insie the unit circle, the inverse of the roots are use to make the moels stationary an invertible. The conence intervals of the an estimates are calculate only when an interior maximum is foun, i.e. if j=^z AR j :99 an : The pile-up eect The probability that parameter estimates of are observe when tting MA() or AR() moels to small ata sets are obtaine by Monte Carlo simulation. Table gives the relative frequencies (with respect to replications) that the estimates ier from + or no more than.35 (cp. the numerical restrictions given above) for series of the length of 25, 5 an. E.g., P( ^ = ) stans for P( ^ 2 [:9965; ]). ** inclue TABLE ** The results for the (), the where the true mean is use to emean the ata, compare to the theoretical results of Cryer an Leolter(98). The an the results are very similar to the theoretical ones in case of the MA() moel with = :9. However, for = :9 the pile-up eect is smaller, especially for the. The, which is inepenent of the mean correction, exhibits small frequencies for = :9, an is comparable to the for = :9. For the AR() processes the pile-up eect is less pronounce. The probability of estimates of the are somewhat larger than those for the. 7

11 3.3 Moels an criteria We will compare four estimators for low orer ARFIMA moels with respect to the mean square error,, of the parameter estimates, the bias, an the percentage that the true parameter lies in the 95% conence interval base on the asymptotic normal istribution. This percentage is enote by, empirical conence level. The values given for the are scale by a factor of. The length of the series consiere is. replications are performe. *** INCLUDE somewhere below FIGURES to an TABLE 2 The moels consiere in the simulation stuy are as follows: AR() with parameter values = :99; :95;... ; :95; :99. The results are summarize in Figure not incluing the estimator. The is neither presente graphically nor in tables (except in Fig. 3) but escribe verbally at the en of this section ue to its limite applicability. MA() with = :; :99; :95;... ; :95; :99; : (Figure 2). FI() with values ranging form :5 to :45; :49 (Figure 3). AR(2) moels with real roots, z ;2, close together as in Zhang(99). For z < the relation =z + :5 = =z 2 is assume, for z > =z :5 = =z 2, with =z = :95; :9;... ; :9; :95. See Figure 4. The same parameter values are implemente in MA(2) moels yieling troughs in the spectrum where the AR moels have peaks (Figure 5). These moels are of interest, since the small sample behavior of the Whittle estimator using tapere ata has essentially been applie only to AR moels. In aition to real roots, we also consier complex roots, z ;2 = r e i, in AR(2) moels (Figure 6). We choose a constant moulus close to the unit circle of r = =:95, an uniformly istribute frequencies over [; ]: = j with j = ;... ;. Both ens, j = ;, of the interval correspon to ouble real roots. Due to the similarity of the behavior of both involve parameter estimates an to space restrictions only the of the secon parameter estimate are given in Fig. 4 to 6. Further, ARMA(,) processes are investigate (Figure 7). The an parameters are chosen to take the values :99; :95;... ; :95; :99. For the list of values is augmente by :. The minimal istance between the AR an MA parameter is chosen to be :5, j j :5. Aitional moels with almost canceling roots are specie as 8

12 pairs ( ; :5) an ( :5; ) for ; = :8; :6;... ; :. The contour lines of the surfaces in Fig. 7.a are rawn to inicate the 5:; :;... levels. (The computations were performe with the triangle contour plot algorithm of Preusser(984).) In orer to accentuate a bias of null the corresponing contour line is rawn fat. The other bias levels are otte inicating levels of :5; :;.... In a similar way the empirical conence levels are marke. The fat lines inicate a level of 95%, the otte ones ; 9; 85;.... Simulation results for ARFIMA(,,) processes are summarize in Figure 8. The values for the fractional integration parameter are chosen between :5 an.45. The autoregressive parameter varies between :95 an.95. However contrary to Fig. 7, the contour levels in the plots are 2:5; 5:;.... Finally, the results for ARFIMA(,,) processes with parameters also ranging from :5 to.45, an from :95 to.95 are given. See Figure. In aition to the graphical representation, the maximal an minimal values of each criterion for moels with 2-imensional parameter spaces, together with the points where they are assume, are collecte in Table 2. Since an estimator may exhibit the largest maximum in a small area but performs excellent for most other moels, while another estimator may be rather ba over the whole class of moels, the overall maximum may be no aequate inicator. So we construct a measure of net avantage of estimator A over estimator B base on the maximal ierences of the surfaces. We ene it as max j A() B ()j max j A() B ()j : B>A A>B A () enotes the of the parameter estimate of obtaine by estimator A. The maximal avantage (measure in ) using estimator A instea of B is so relate to the maximal avantage of estimator B. If our measure is positive, estimator A is to be preferre over estimator B. If it is negative, B is to be preferre. Table 3 summarizes the results for invertible ARMA(,), ARFIMA(,,) an ARFIMA(,,) moels. 3.4 Small sample behavior of the estimators Our numerical results compare - as far as values are available - well to previous stuies like Zhang(99) an Cheung an Diebol(994). However, they ier to some extent from Ansley an Newbol(98). They emean the ata by the population mean while 9

13 we use the sample mean as is commonly one leaing to a more or less pronounce asymmetric behavior of the time omain ML estimates at parameter values associate with real roots close to + on one han, an on the other. In the following we survey the small sample properties of pairs of estimators by rst stating the main properties an then backing up the conclusion by referencing to the gures an tables provie. We compare an, an, give a verbal escription of the, an iscuss the eects of tapering the ata. For non fractionally integrate moels the turns out to be either equivalent to or more favorable than the. In case of AR() moels with negative the behavior of an are essentially ientical. For larger positive the an bias of the estimates are smaller, an the is closer to the theoretical 95% level. The same relation hols for the MA() moels (with the exception of close to +). In case of the AR(2) (Fig. 4 an 6) an the MA(2) moels (Fig. 5) ierences between both estimators are not visible. For the ARMA(,) moel the minimal of the is the same as that of the (.6). The maximum of 54.67, however, is clearly below the corresponing value of the, The criterion of the net avantage between an (see Tab. 3) rates both estimators as essentially equivalent. For fractionally integrate moels the is clearly ominate by the. This hols particularly for mixe autoregressive or moving average fractionally integrate moels. Compare Fig. 3, 8 to an Tab. 2. The systematic negative bias in the estimate of the is enlarge by the inclusion of the AR an MA parameter. E.g., in case of the ARFIMA(,,) moel with true values of (; ) = (:5; :2) the meian estimate moel is me( ) = :5 an me( )=.426! The associate true an estimate spectral ensities are plotte in Figure 9. It turns out that the cannot etect the true sign of the parameter for a rather large set of moels. Those moels are marke by crosses in the bias plots. Compare both Fig. 8.b an Fig..b. The large bias of the estimate is reecte in an. The, on the other han, is not aecte by that systematic bias an has goo an acceptable properties. The criterion of net avantage with respect to the conrms our conclusions. See Tab. 3. The, the Whittle likelihoo applie to tapere ata, exhibits for \well-behave"

14 moels slightly larger than the. However, for more \icult" moels this isavantage seems to be compensate suciently. The turns out to be a reliable estimator. The favorable properties of the o not come into eect for the single parameter moels AR() an FI(). In those cases the is larger than that of the over the whole parameter region. But bias an properties are essentially equivalent. See Fig. an 3. For most values this relation also hols in the MA() moels. However close to = :,, bias an are more favorable with respect to the (Fig. 2). Similarly, as in the MA() case the ominates the close to =z = :95 for the AR(2) moels in Fig. 4, close to =z = :95 for the MA(2) moels in Fig. 5, an close to = for the AR(2) moels with conjugate complex roots in Fig. 6. For ARMA(,) moels the avantage of the is evient (Fig. 7 an Tab. 2, 3). Moels with almost canceling polynomials can be estimate with the more precisely. The sprea of the bias an the maximal are consierably smaller for the for both ^ an ^, than for the. The net avantage as given in Tab. 3 is also in favor of the. The of the for ^ at the true parameters ( ; ) = ( :99; :) is extremely low (6.9), while the minimal value for the is Fig. 8.c also shows that the conence intervals base on or estimates for values close to + are too small. (We have also investigate the sensitivity of our conclusion ue to the inclusion of moels with possibly too strong canceling eects in relation to the length of the series. However, the avantage of the remains if the moel class is restricte to j j :2. In this case the maximal values of an are rather the same, but net avantage, bias an are still in favor of the. The corresponing values are not tabulate.) Accoring to visual inspection of Fig. 8, minimal an maximal, the net avantage with respect to the (cp. Tab. 3), an sprea of the bias (see Tab. 2), the is to be preferre for the ARFIMA(,,) moels. However, its for may be rather low for moels close to (; ) = ( :5; :95). On the contrary, in case of the ARFIMA(,,) moels all criteria apart from the net avantage are in favor of the. The estimates may exhibit serious eciencies when the roots of the moels are \close" to real unit roots, both for AR an MA polynomials. Unacceptable high values are observe together with large biases in relative large parameter ranges. So, the

15 cannot be recommene in general, an a etaile graphical presentation is omitte. The most important simulation results are ocumente in the following in a verbal way. The efects of the may harly be observe for AR() moels. For MA() moels, AR(2) an MA(2) moels with real roots, however, the curves aopt a pronounce trough shape with a rather at bottom. In the AR(2) moels with complex roots a U- shape is observe (with a minimal value at = =2 which competes well with the ). The surface for ^ of the ARMA(,) moels exhibits a W-like prole with the rige generate by the moels with almost canceling roots in the mile. (The surface for ^ is not aecte an similar to the one observe for the.) The estimate of the fractional integration parameter, contrary to the, oes not exhibit a systematic bias. This is most evient for the FI() moels, where the is the secon best estimator being only slightly inferior to the, an more favorable than the. See Fig. 3. Incluing an AR parameter in the FI() moel the of both parameters become rather high at (; ) = (:45; :95), while for the ARFIMA (,,) the overall properties are inferior to the but compete with the. Tapering the ata increases the but not the bias for \well-behave" moels an lowers both, an bias, impressively for all problematic moels investigate. 3.5 An explanation for the bias in the estimate of The potential negative bias in the estimate obtaine by the, in contrast to the where no systematic bias is observe, seems worth to be investigate more closely. A heuristic explanation for this behavior is oere below for fractionally integrate processes with >. We observe that both estimators ier in the treatment of the mean, the frequency zero in the spectral representation respectively. While frequency zero is exclue in the estimation explicitly, it is implicitly inclue in the through the autocovariance function (h). (h) = Z f() e i h () where the covariance matrix is = () = [(ji jj)] i;j with = ( ;... ; p ; ; ;... ; q ). In case of emeane ata the perioogram as estimator of the spectrum is null at frequency zero inepenently on the process. However, for > the spectrum is innite at zero. As the parameter essentially escribes only the slope of the spectrum close to frequency 2

16 zero, it is very sensitive to changes in the low frequency region (but insensitive to high frequency eects). Now consier the replacement of f in () by the perioogram for emeane ata. Then the estimate of will try to moel the upwar slope of the true spectrum for low frequencies, but will have to take into account the value of null at zero. So, a negative bias will result. Not surprisingly, the negative bias in case of the ARFIMA(,,) an (,,) moels may be consierably larger than in pure FI moels. The aitional ARMA parameter (which moels the spectrum at low an high frequencies) oers more exibility to to capture the suen ecrease at frequency zero. The consequence is also a consierable bias in ^, ^ respectively, ue to a compensating eect in the estimate spectrum at low frequencies (cp. Fig. 9). Our explanation is in line with the negative bias in the sample autocovariance function as observe by Newbol an Agiakloglou(993). It is also compatible with the nings of Cheung an Diebol(994) that the exact maximum likelihoo estimator for ata correcte with the true mean yiels unbiase estimates. 4 Summary We investigate the pile-up eect for the exact maximum likelihoo metho, the moi- e prole likelihoo, the Whittle estimator an the Whittle estimator for tapere ata. Some conclusions are: The pile-up eect of the applie to emeane ata is somewhat smaller than the theoretical values for known mean. Contrary to the, the exhibits the pile-up eect also for the AR() moels, but to a lesser extent. An extensive simulation stuy is performe to investigate the properties of the estimators in low orer ARMA an ARFIMA moels. The length of the series is. Despite the asymptotic equivalence of the estimators we n that their small sample properties may ier substantially. The exact maximum likelihoo exhibits a rather ba performance in mixe ARFI or FIMA moels. The estimates ten to be seriously negatively biase leaing to large mean square errors an low empirical conence levels. Also, for ARMA(,) moels with almost canceling roots the use of the may lea to extremely large mean square errors an biases. The moie prole likelihoo exhibits for ARMA moels slightly better properties than the exact maximum likelihoo estimator, 3

17 an ominates it clearly for fractionally integrate ones. The Whittle likelihoo, while yieling no systematically biase estimates, has serious eciencies for large parameter ranges especially \close" to real roots of +=, an so cannot be recommene in general. Base on a comparison of the an we oer a heuristic explanation for the bias in the estimates of in long memory moels. The Whittle likelihoo with tapere ata performs well not only for AR moels but also for MA, ARMA(,) an fractionally integrate moels. It turns out to be an overall reliable estimator. The small losses in performance in case of \well-behave" moels seem to be compensate suciently in more \icult" moels. Its computational simplicity is also attractive. The computationally more emaning alternative with certain avantages is the moie prole likelihoo. Caution is requeste for ARMA(,) moels with almost canceling roots in general, an, in particular, in case of the an the for inference in the vicinity of a moving average root of +. 4

18 References An, S., P. Bloomel an S. Pantula, 992, Asymptotic properties of the MLE in fractional ARIMA moels, Institute of Statistics Mimeograph Series No. 2228, North Carolina State University. An, S. an P. Bloomel, 993, Cox an Rei's moication in regression moels with correlate errors, Department of statistics, North Carolina State University, Raleigh. Anerson, T.W. an A. Takemura, 986, Why o noninvertible estimate moving averages occur?, Journal of Times Series Analysis, 7, Ansley, C.F. an P. Newbol, 98, Finite sample properties of estimators for autoregressive moving average moels, Journal of Econometrics, 3, Bloomel, P., 985, On series representations for linear preictors, The Annals of Probability, 3, Boes, D.C., R.A. Davis an S.N. Gupta, 989, Parameter estimation in low orer fractionally ierence ARMA processes, Stochastic Hyrology an Hyraulics, 3, 97-. Brockwell, P.J. an R.A. Davis, 99, Time series: Theory an methos (Springer, New York). Cheung, Y.-W. an F.X. Diebol, 994, On maximum-likelihoo estimation of the ierencing parameter of fractionally-integrate noise with unknown mean, Journal of Econometrics, 62, Cox, D.R. an N. Rei, 987, Parameter orthogonality an approximate conitional inference (with iscussion), Journal of the Royal Statistical Society Series B, 49, -39. Cryer, J. an J. Leolter, 98, Small-sample properties of the maximum likelihoo estimator in the rst-orer moving average moel, Biometrika, 68, Dahlhaus, R., 988, Small sample eects in time series analysis: A new asymptotic theory an a new estimate, The Annals of Statistics, 6, Dahlhaus, R., 989, Ecient parameter estimation for self-similar processes, The Annals of Statistics, 7, Dzhaparize, K., 986, Parameter estimation an hypothesis testing in spectral analysis of stationary time series (Springer, New York). Fox, R. an M.S. Taqqu, 986, Large-sample properties of parameter estimates for strongly epenent stationary gaussian time series, The Annals of Statistics, 4, Giraitis, L. an D. Surgailis, 99, A central limit theorem for quaratic forms in strongly epenent linear variables an its application to asymptotic normality of Whittle's estimate, Probability Theory an Relate Fiels, 86,

19 Hauser, M.A., 995, Long range epenence in international output series: A reexamination, (Worl Congress of the Econometric Society, Tokyo) Hoermann, W. an G. Deringer, 99, The ACR metho for generating normal ranom variables, OR Spektrum, 2, Hurvich, C.M., 988, A mean square criterion for time series ata winows, Biometrika, 75, Kang, H.-J., 987, The tapering estimation of the rst-orer autoregressive parameters, Biometrika, 74, Li, W.K. an A.I. McLeo, 986, Fractional time series moelling, Biometrika, 73, Matsumoto, M. an Y. Kurita, 994, Twiste GFSR generators, II, ACM Transactions on Moelling an Computer Simulation, 4, 994, Newbol, P. an Ch. Agiakloglou, 993, Bias in the sample autocorrelation of fractional noise, Biometrika, 8, Parzen, E., 983, Autoregressive spectral estimation. In Hanbook of Statistics (Brillinger, D.R. an Krishnaiah, P.R., es.) 3, , (North-Hollan, Amsteram). Priestley, M.B., 98, Spectral analysis of time series, Volume : Univariate Series (Acaemic Press, Lonon). Preusser, A., 984, Triangle contour plotting, ACM-Trans. Math. Software,, (Algorithm 626 Collecte algorithms from ACM). Pukkila, T., an H. Nyquist, 985, On the frequency omain estimation of the innovation variance of a stationary univariate time series, Biometrika, 72, Robinson, P.M., 99, Time series with strong epenence, Paper presente at the Worl Congress of the Econometric Society in Barcelona, August 99. Sowell, F., 992a, Maximum likelihoo estimation of stationary univariate fractionally integrate time series moels, Journal of Econometrics, 53, Sowell, F., 992b, Moeling long-run behavior with the fractional ARIMA moel, Journal of Monetary Economics, 29, Whittle, P., 962, Gaussian estimation in stationary time series, Bulletin of the International Institute of Statistics, 39, Zhang, H.-C., 99, Reuction of the asymptotic bias of autoregressive an spectral estimators by tapering, Journal of Time Series Analysis, 3,

20 TABLE : Pile-up eect in MA() an AR() moels for n=25, 5,. The relative frequency that ^ ( = ; ) is in the one-sie interval [ :; :9965] or [.9965,.]. The estimators are,,,, an (), the exact Gaussian ML with ata correcte for the true mean. The theoretical values obtaine by Cryer an Leolter (98) are enote by th. The number of replications for our simulations is. P(^ = ) P(^ = ) MA(): = :9 = :9 n th () AR(): = :9 = :9 n Remark: Distances from the theoretical values which excee.3,.2 or.9 epening on the theoretical values.53,.333 an.36 are signicant at the % level. 7

21 TABLE 2: Minima an maxima of, bias an empirical conence level for the estimators,, in ARMA(,), ARFIMA(,,) an ARFIMA(,,) moels, an where they are assume. ARMA(,) ^ ^ (, ) min (, ) max (, ) min (, ) max ARFIMA(,,) ^ (, ) min (, ) max (, ) min (, ) max ARFIMA(,,) ^ (, ) min (, ) max (, ) min (, ) max

22 TABLE 3: The measure of net avantage in of estimator A over estimator B for invertible ARMA(,), ARFIMA(,,) an ARFIMA(,,) moels: max j A() B ()j max j A() B ()j B>A A>B with A () as the of ^ using the estimator A. The estimators consiere are:, an. ARMA(,) = = Q QQ A B Q Q QQ A Q ARFIMA(,,) = = B Q QQ A Q ARFIMA(,,) = = B Remark: A positive sign inicates an avantage of estimator A over estimator B, a negative an avantage of estimator B over A. 9

23 * ^ ^ =z ^ 2= 4= ^ 6= 8= ^ ^ ^ FIGURE : Mean ^ square error, bias an empirical conence level for the AR() moels. PSfrag replacements ^ ^ ^ ^ ^ =z ^.5 ^ 2= ^ 4= ^ 6= ^ 8= * ^ ^ ^ ^ ^ ^ ^ ^ ^.5 =z ^ 2= 4= ^ ^ 6= 8= ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^. ^ * ^ ^ ^ ^ =z ^ 2= 4= ^ -. 6= 8= * ^ ^ ^ ^ ^ ^ ^ ^ 9 ^ ^ ^ ^ ^ ^ *

24 * =z 2= 4= 6= 8= ^ ^ FIGURE 2: Mean square error, bias an empirical conence level for the MA() moels. ^ ^ ^ ^.5 ^ ^ ^ ^ PSfrag replacements ^ ^.5 =z ^ 2= 4= ^ 6= 8= ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ PSfrag replacements ^ ^ =z ^ 2= 4= ^ 6= 8= * ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ *

25 * =z 2= 4= 6= 8= ^ ^ FIGURE 3: Mean square error, bias an empirical conence level for the FI() moels. ^ ^ ^ ^.5 ^ ^ ^ ^ PSfrag replacements ^ ^.5 =z ^ 2= 4= ^ 6= 8= ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ PSfrag replacements ^ ^ =z ^ 2= 4= ^ 6= 8= * ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ *

26 PSfrag replacements * * * 2= 4= 6= 8= FIGURE 4: Mean square ^ error for the AR(2) moels with real roots z;2 with =z = :95; : : : ; :5, ^ an =z2 = =z + :5; =z = :5; : : : ; :95, an =z2 = =z :5. ^ ^ ^ ^ ^ 2 ^ ^ ^ ^ ^ ^ ^ PSfrag replacements =z 2= 4= FIGURE 6= 5: Mean square error for the MA(2) moels with real roots z;2 with 8= =z = :; ^ :95; : : : ; :5, an =z2 = =z + :5; =z = :5; : : : ; :95; :, an ^ =z2 = =z :5. ^ ^ ^ ^ ^ 2 ^ ^ ^ ^ ^ ^ ^ PSfrag replacements =z FIGURE 6: Mean square ^ error for the AR(2) moels with complex roots z;2 = re i with =r = :95 ^ an = ; ; 2 ; : : : ; 9 ;. ^ ^ ^ ^ ^ 2 ^ ^ ^ ^ ^ ^ ^.5.5 =z 2= 4= 6= 8= 23

27 =z ^ * IAS ^ ^ ^ ^ =z ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ =z ^ ^ * IAS ^ ^ ^ ^ ^ ^ ^ =z ^ ^ ^ IAS ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ * IAS ^ ^ ^ ^ =z ^ IAS ^ ^ ents * * -.5 ^ ^ ^ ^ ^ ^ * =z ^ 2= * 4= 6= 8= ^ ^ ^ FIGURE 7: Results for the ARMA(,) ^ moels. =z ^ 2= ^ FIGURE 7.a: Mean square error for the 4= 6= ARMA(,) ^ moels. 8= ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ PSfrag replacements ^ ^ * ^ ^ =z ^ 7 2= ^ 6 5 * 4= 6= ^ 8= 4 ^ 3 2 ^ ^ ^ ^ ^ -.99 ^ -.5 =z ^ - 2= ^ = 6= ^.99 8= ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ * ^ ^ ^ 7 ^ 6 5 * ^ 4 ^ 3 2 ^ ^ -.99 ^ -.5 =z - 2= = 6=.99 8= ^ ^ ^ ^ ^ ^ ^ *.5.5 ^.99 ^.99 ^ ^ ^ ^ ^ ^ ^ IAS ^ ^ ^ ^ ^ -.99 ^

28 IAS ^ ^ ^ ^ ^ ^ ^ =z ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ * IAS ^ ^ ^ ^ ^ ^ ^ ^ ^ =z ^ ^ * IAS ^ ^ ^ ^ ^ ^ ^ =z ^ ^ ^ * ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ * IAS ^ ^ ^ ^ =z ^ * ^ ^ ^ ^ ^ ^ ^ ^ 6= 8= ^ ^ ^ ^ ^ ^ ^ =z ^ 2= ^ ^ 4= FIGURE 7.b: Bias for the ARMA(,) ^ 6= ^ moels. 8= ^ ^ ^ ^ ^ ^ * ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ =z ^.6 2= ^.4 * 4= 6= ^.2 8= ^ ^ -.2 ^ ^ -.4 ^ ^ -.99 ^ -.5 =z ^ - 2= ^ * 4= 6= ^.99 8= ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^.6 ^.4 * ^.2 ^ -.2 ^ -.4 ^ -.99 ^ -.5 =z - 2= * 4= 6=.99 8= ^ ^ ^ ^ ^ ^ ^.5.5 ^.99 ^.99 ^ ^ ^ ^ ^ ^ ^ * ^ ^ ^ ^ ^ -.99 ^ -.5 *

29 IAS ^ ^ ^ ^ ^ ^ ^ =z ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ * IAS ^ ^ ^ ^ ^ ^ ^ ^ ^ =z ^ ^ * IAS ^ ^ ^ ^ ^ ^ ^ =z ^ ^ ^ * IAS ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ * IAS ^ ^ ^ ^ =z ^ * IAS ^ ^ ^ ^ ^ ^ ^ ^ 6= 8= ^ ^ ^ ^ ^ ^ ^ =z ^ 2= ^ ^ 4= FIGURE 7.c: Empirical conence level for 6= the ^ ARMA(,) ^ moels. 8= ^ ^ ^ ^ ^ ^ * ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ =z ^ 2= ^ 9 * 4= 6= ^ 8= 8 ^ 7 ^ ^ 6 ^ ^ ^ -.99 ^ -.5 =z ^ - 2= ^ * 4= 6= ^.99 8= ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ 9 * ^ 8 ^ 7 ^ 6 ^ -.99 ^ -.5 =z - 2= * 4= 6=.99 8= ^ ^ ^ ^ ^ ^ ^.5.5 ^.99 ^.99 ^ ^ ^ ^ ^ ^ ^ * IAS ^ ^ ^ ^ ^ -.99 ^ -.5 *

30 =z ^ ^ ^ * IAS ^ ^ ^ ^ ^ ^ ^ ^ =z ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ * IAS ^ ^ ^ ^ ^ ^ ^ ^ ^ =z ^ ^ * IAS ^ ^ ^ ^ ^ ^ ^ =z ^ ^ ^ IAS ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ * IAS ^ ^ ^ ^ =z ^ IAS ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ IAS * * * =z ^ ^ 2= ^ * 4= 6= ^ 8= ^ ^ ^ ^ ^ FIGURE 8: Results for the ARFIMA(,,) ^ ^ moels. =z ^ 2= ^ ^ FIGURE 8.a: Mean square error for the ARFIMA(,,) 4= ^ 6= ^ moels. 8= ^ ^ ^ ^ ^ ^ * ^ ^ ^ ^ ^ ^ ^ * ^ ^ =z ^ 25 2= ^ 2 * 4= 6= ^ 5 8= ^ ^ 5 ^ ^ ^ ^ =z ^ ^ = ^ 4= = ^ = ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ * ^ ^ ^ 25 ^ 2 * ^ 5 ^ 5 ^ ^ =z ^ = 4= = = ^ ^ ^ ^ ^ ^ * ^ 25 ^ 2 ^ 5 ^ 5 ^ ^ ^ ^ ^ ^

31 IAS 6= ^ 8= ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ =z ^ =z ^ ^ ^ 2= ^ ^ ^ FIGURE 8.b: Bias for the ARFIMA(,,) moels. A cross 4= inicates ^ ^ 6= ^ a wrong sign of the average. ^ ^ 8= ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ * ^ ^ * IAS ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ =z ^.4.4 =z ^ ^.2 ^.2 2= * IAS ^ * 4= 6= ^ 8= ^ -.2 ^ -.2 ^ ^ -.4 ^ ^ -.4 ^ ^ ^ ^ ^ ^ =z ^ ^ ^ =z ^ ^ 2= ^ * * 4= ^ = ^.25 ^ = ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^.4 ^.4 ^.2 ^.2 * IAS ^ * ^ ^ -.2 ^ -.2 ^ -.4 ^ -.4 ^ ^ =z ^ ^ =z = * * 4= = = ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ * ^ ^ ^ ^ ^ ^ ^ ^.95 ^ *

32 IAS ^ ^ ^ ^ ^ ^ ^ =z ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ * IAS ^ ^ ^ ^ ^ ^ ^ ^ ^ =z ^ ^ * IAS ^ ^ ^ ^ ^ ^ ^ =z ^ ^ ^ * IAS ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ * IAS ^ ^ ^ ^ =z ^ * IAS ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ * IAS 6= 8= ^ ^ ^ ^ ^ ^ ^ =z ^ 2= ^ ^ 4= FIGURE 8.c: Empirical conence level for the 6= ARFIMA(,,) ^ ^ moels. 8= ^ ^ ^ ^ ^ ^ * ^ ^ ^ ^ ^ ^ ^ ^ ^ =z ^ 9 2= ^ 9 8 * 4= 6= ^ 8 8= 7 ^ 7 6 ^ ^ 6 ^ ^ ^ =z ^ ^ = ^ * 4= = ^ = ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ 9 ^ 9 8 * ^ 8 7 ^ 7 6 ^ 6 ^ =z ^ = * 4= = = ^ ^ ^ ^ ^ ^ ^ 9 ^ 9 8 ^ 8 7 ^ 7 6 ^ 6 ^ ^ * ^ ^ ^

33 ^ ^ ^ ^ ^ FIGURE 9: Spectrum of the ARFIMA(,,) moel with = :5, an = :2, an spectrum of the ^ meian moel estimate by with me( ) = :5 an me(^) = :426 (ashe line). * =z 2= 4= 6= 8= 7 ^ ^ 6 ^ ^ ^ 5 ^ ^ 4 ^ ^ ^ ^ ^ ^ ^ *

34 =z ^ ^ ^ * IAS ^ ^ ^ ^ ^ ^ ^ ^ =z ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ * IAS ^ ^ ^ ^ ^ ^ ^ ^ ^ =z ^ ^ * IAS ^ ^ ^ ^ ^ ^ ^ =z ^ ^ ^ IAS ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ * IAS ^ ^ ^ ^ =z ^ IAS ^ ^ ^ ^ ^ ^ ^ ^ * * * =z ^ ^ 2= ^ * 4= 6= ^ 8= ^ ^ ^ ^ ^ FIGURE : Results for the ARFIMA(,,) ^ ^ moels. =z ^ 2= ^ ^ FIGURE.a: Mean square error for the ARFIMA(,,) 4= ^ 6= ^ moels. 8= ^ ^ ^ ^ ^ ^ * ^ ^ ^ ^ ^ ^ ^ * ^ ^ =z ^ 25 2= ^ 2 * 4= 6= ^ 5 8= ^ ^ 5 ^ ^ ^ ^ =z ^ ^ = ^ 4= = ^ = ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ * ^ ^ ^ 25 ^ 2 * ^ 5 ^ 5 ^ ^ =z ^ = 4= = = ^ ^ ^ ^ ^ ^ ^ * ^ ^ ^ ^ ^ ^ ^ ^ ^ IAS ^ ^ ^ ^ ^.95 ^

35 IAS 6= ^ 8= ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ =z ^ =z ^ ^ ^ 2= ^ ^ ^ FIGURE.b: Bias for the ARFIMA(,,) moels. A cross 4= inicates ^ ^ 6= ^ a wrong sign of the average. ^ ^ 8= ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ * ^ ^ * IAS ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ =z ^.4.4 =z ^ ^.2 ^.2 2= * IAS ^ * 4= 6= ^ 8= ^ -.2 ^ -.2 ^ ^ -.4 ^ ^ -.4 ^ ^ ^ ^ ^ ^ =z ^ ^ ^ =z ^ ^ 2= ^ * * 4= ^ = ^.25 ^ = ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^.4 ^.4 ^.2 ^.2 * IAS ^ * ^ ^ -.2 ^ -.2 ^ -.4 ^ -.4 ^ ^ =z ^ ^ =z = * * 4= = = ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ * ^ ^ ^ ^ ^.95 ^ *

36 IAS ^ ^ ^ ^ ^ ^ ^ =z ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ * IAS ^ ^ ^ ^ ^ ^ ^ ^ ^ =z ^ ^ * IAS ^ ^ ^ ^ ^ ^ ^ =z ^ ^ ^ * IAS ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ * IAS ^ ^ ^ ^ =z ^ * IAS ^ ^ ^ ^ ^ ^ ^ ^ 6= 8= ^ ^ ^ ^ ^ ^ ^ =z ^ 2= ^ ^ 4= FIGURE.c: Empirical conence level for the ^ 6= ARFIMA(,,) ^ moels. 8= ^ ^ ^ ^ ^ ^ * ^ ^ ^ ^ ^ ^ ^ ^ ^ =z ^ 9 2= ^ 9 8 * 4= 6= ^ 8 8= 7 ^ 7 6 ^ ^ 6 ^ ^ ^ =z ^ ^ = ^ * 4= = ^ = ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ 9 ^ 9 8 * ^ 8 7 ^ 7 6 ^ 6 ^ =z ^ = * 4= = = ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ * IAS ^ ^ ^ ^ ^.95 ^ *

Chapter 6: Energy-Momentum Tensors

Chapter 6: Energy-Momentum Tensors 49 Chapter 6: Energy-Momentum Tensors This chapter outlines the general theory of energy an momentum conservation in terms of energy-momentum tensors, then applies these ieas to the case of Bohm's moel.

More information

1 Introuction In the past few years there has been renewe interest in the nerson impurity moel. This moel was originally propose by nerson [2], for a

1 Introuction In the past few years there has been renewe interest in the nerson impurity moel. This moel was originally propose by nerson [2], for a Theory of the nerson impurity moel: The Schrieer{Wol transformation re{examine Stefan K. Kehrein 1 an nreas Mielke 2 Institut fur Theoretische Physik, uprecht{karls{universitat, D{69120 Heielberg, F..

More information

A Modification of the Jarque-Bera Test. for Normality

A Modification of the Jarque-Bera Test. for Normality Int. J. Contemp. Math. Sciences, Vol. 8, 01, no. 17, 84-85 HIKARI Lt, www.m-hikari.com http://x.oi.org/10.1988/ijcms.01.9106 A Moification of the Jarque-Bera Test for Normality Moawa El-Fallah Ab El-Salam

More information

Qubit channels that achieve capacity with two states

Qubit channels that achieve capacity with two states Qubit channels that achieve capacity with two states Dominic W. Berry Department of Physics, The University of Queenslan, Brisbane, Queenslan 4072, Australia Receive 22 December 2004; publishe 22 March

More information

Improving Estimation Accuracy in Nonrandomized Response Questioning Methods by Multiple Answers

Improving Estimation Accuracy in Nonrandomized Response Questioning Methods by Multiple Answers International Journal of Statistics an Probability; Vol 6, No 5; September 207 ISSN 927-7032 E-ISSN 927-7040 Publishe by Canaian Center of Science an Eucation Improving Estimation Accuracy in Nonranomize

More information

A representation theory for a class of vector autoregressive models for fractional processes

A representation theory for a class of vector autoregressive models for fractional processes A representation theory for a class of vector autoregressive moels for fractional processes Søren Johansen Department of Applie Mathematics an Statistics, University of Copenhagen November 2006 Abstract

More information

the solution of ()-(), an ecient numerical treatment requires variable steps. An alternative approach is to apply a time transformation of the form t

the solution of ()-(), an ecient numerical treatment requires variable steps. An alternative approach is to apply a time transformation of the form t Asymptotic Error Analysis of the Aaptive Verlet Metho Stephane Cirilli, Ernst Hairer Beneict Leimkuhler y May 3, 999 Abstract The Aaptive Verlet metho [7] an variants [6] are time-reversible schemes for

More information

Least-Squares Regression on Sparse Spaces

Least-Squares Regression on Sparse Spaces Least-Squares Regression on Sparse Spaces Yuri Grinberg, Mahi Milani Far, Joelle Pineau School of Computer Science McGill University Montreal, Canaa {ygrinb,mmilan1,jpineau}@cs.mcgill.ca 1 Introuction

More information

Generalization of the persistent random walk to dimensions greater than 1

Generalization of the persistent random walk to dimensions greater than 1 PHYSICAL REVIEW E VOLUME 58, NUMBER 6 DECEMBER 1998 Generalization of the persistent ranom walk to imensions greater than 1 Marián Boguñá, Josep M. Porrà, an Jaume Masoliver Departament e Física Fonamental,

More information

IERCU. Institute of Economic Research, Chuo University 50th Anniversary Special Issues. Discussion Paper No.210

IERCU. Institute of Economic Research, Chuo University 50th Anniversary Special Issues. Discussion Paper No.210 IERCU Institute of Economic Research, Chuo University 50th Anniversary Special Issues Discussion Paper No.210 Discrete an Continuous Dynamics in Nonlinear Monopolies Akio Matsumoto Chuo University Ferenc

More information

Survey Sampling. 1 Design-based Inference. Kosuke Imai Department of Politics, Princeton University. February 19, 2013

Survey Sampling. 1 Design-based Inference. Kosuke Imai Department of Politics, Princeton University. February 19, 2013 Survey Sampling Kosuke Imai Department of Politics, Princeton University February 19, 2013 Survey sampling is one of the most commonly use ata collection methos for social scientists. We begin by escribing

More information

Spurious Significance of Treatment Effects in Overfitted Fixed Effect Models Albrecht Ritschl 1 LSE and CEPR. March 2009

Spurious Significance of Treatment Effects in Overfitted Fixed Effect Models Albrecht Ritschl 1 LSE and CEPR. March 2009 Spurious Significance of reatment Effects in Overfitte Fixe Effect Moels Albrecht Ritschl LSE an CEPR March 2009 Introuction Evaluating subsample means across groups an time perios is common in panel stuies

More information

The Role of Models in Model-Assisted and Model- Dependent Estimation for Domains and Small Areas

The Role of Models in Model-Assisted and Model- Dependent Estimation for Domains and Small Areas The Role of Moels in Moel-Assiste an Moel- Depenent Estimation for Domains an Small Areas Risto Lehtonen University of Helsini Mio Myrsylä University of Pennsylvania Carl-Eri Särnal University of Montreal

More information

A simple model for the small-strain behaviour of soils

A simple model for the small-strain behaviour of soils A simple moel for the small-strain behaviour of soils José Jorge Naer Department of Structural an Geotechnical ngineering, Polytechnic School, University of São Paulo 05508-900, São Paulo, Brazil, e-mail:

More information

Research Article When Inflation Causes No Increase in Claim Amounts

Research Article When Inflation Causes No Increase in Claim Amounts Probability an Statistics Volume 2009, Article ID 943926, 10 pages oi:10.1155/2009/943926 Research Article When Inflation Causes No Increase in Claim Amounts Vytaras Brazauskas, 1 Bruce L. Jones, 2 an

More information

Table of Common Derivatives By David Abraham

Table of Common Derivatives By David Abraham Prouct an Quotient Rules: Table of Common Derivatives By Davi Abraham [ f ( g( ] = [ f ( ] g( + f ( [ g( ] f ( = g( [ f ( ] g( g( f ( [ g( ] Trigonometric Functions: sin( = cos( cos( = sin( tan( = sec

More information

Time-of-Arrival Estimation in Non-Line-Of-Sight Environments

Time-of-Arrival Estimation in Non-Line-Of-Sight Environments 2 Conference on Information Sciences an Systems, The Johns Hopkins University, March 2, 2 Time-of-Arrival Estimation in Non-Line-Of-Sight Environments Sinan Gezici, Hisashi Kobayashi an H. Vincent Poor

More information

Lower Bounds for the Smoothed Number of Pareto optimal Solutions

Lower Bounds for the Smoothed Number of Pareto optimal Solutions Lower Bouns for the Smoothe Number of Pareto optimal Solutions Tobias Brunsch an Heiko Röglin Department of Computer Science, University of Bonn, Germany brunsch@cs.uni-bonn.e, heiko@roeglin.org Abstract.

More information

Influence of weight initialization on multilayer perceptron performance

Influence of weight initialization on multilayer perceptron performance Influence of weight initialization on multilayer perceptron performance M. Karouia (1,2) T. Denœux (1) R. Lengellé (1) (1) Université e Compiègne U.R.A. CNRS 817 Heuiasyc BP 649 - F-66 Compiègne ceex -

More information

arxiv: v1 [hep-lat] 19 Nov 2013

arxiv: v1 [hep-lat] 19 Nov 2013 HU-EP-13/69 SFB/CPP-13-98 DESY 13-225 Applicability of Quasi-Monte Carlo for lattice systems arxiv:1311.4726v1 [hep-lat] 19 ov 2013, a,b Tobias Hartung, c Karl Jansen, b Hernan Leovey, Anreas Griewank

More information

Problems Governed by PDE. Shlomo Ta'asan. Carnegie Mellon University. and. Abstract

Problems Governed by PDE. Shlomo Ta'asan. Carnegie Mellon University. and. Abstract Pseuo-Time Methos for Constraine Optimization Problems Governe by PDE Shlomo Ta'asan Carnegie Mellon University an Institute for Computer Applications in Science an Engineering Abstract In this paper we

More information

Thermal conductivity of graded composites: Numerical simulations and an effective medium approximation

Thermal conductivity of graded composites: Numerical simulations and an effective medium approximation JOURNAL OF MATERIALS SCIENCE 34 (999)5497 5503 Thermal conuctivity of grae composites: Numerical simulations an an effective meium approximation P. M. HUI Department of Physics, The Chinese University

More information

IPA Derivatives for Make-to-Stock Production-Inventory Systems With Backorders Under the (R,r) Policy

IPA Derivatives for Make-to-Stock Production-Inventory Systems With Backorders Under the (R,r) Policy IPA Derivatives for Make-to-Stock Prouction-Inventory Systems With Backorers Uner the (Rr) Policy Yihong Fan a Benamin Melame b Yao Zhao c Yorai Wari Abstract This paper aresses Infinitesimal Perturbation

More information

Estimation of the Maximum Domination Value in Multi-Dimensional Data Sets

Estimation of the Maximum Domination Value in Multi-Dimensional Data Sets Proceeings of the 4th East-European Conference on Avances in Databases an Information Systems ADBIS) 200 Estimation of the Maximum Domination Value in Multi-Dimensional Data Sets Eleftherios Tiakas, Apostolos.

More information

This module is part of the. Memobust Handbook. on Methodology of Modern Business Statistics

This module is part of the. Memobust Handbook. on Methodology of Modern Business Statistics This moule is part of the Memobust Hanbook on Methoology of Moern Business Statistics 26 March 2014 Metho: Balance Sampling for Multi-Way Stratification Contents General section... 3 1. Summary... 3 2.

More information

Web-Based Technical Appendix: Multi-Product Firms and Trade Liberalization

Web-Based Technical Appendix: Multi-Product Firms and Trade Liberalization Web-Base Technical Appeni: Multi-Prouct Firms an Trae Liberalization Anrew B. Bernar Tuck School of Business at Dartmouth & NBER Stephen J. Reing LSE, Yale School of Management & CEPR Peter K. Schott Yale

More information

Logarithmic spurious regressions

Logarithmic spurious regressions Logarithmic spurious regressions Robert M. e Jong Michigan State University February 5, 22 Abstract Spurious regressions, i.e. regressions in which an integrate process is regresse on another integrate

More information

LATTICE-BASED D-OPTIMUM DESIGN FOR FOURIER REGRESSION

LATTICE-BASED D-OPTIMUM DESIGN FOR FOURIER REGRESSION The Annals of Statistics 1997, Vol. 25, No. 6, 2313 2327 LATTICE-BASED D-OPTIMUM DESIGN FOR FOURIER REGRESSION By Eva Riccomagno, 1 Rainer Schwabe 2 an Henry P. Wynn 1 University of Warwick, Technische

More information

Quantum Mechanics in Three Dimensions

Quantum Mechanics in Three Dimensions Physics 342 Lecture 20 Quantum Mechanics in Three Dimensions Lecture 20 Physics 342 Quantum Mechanics I Monay, March 24th, 2008 We begin our spherical solutions with the simplest possible case zero potential.

More information

THE VAN KAMPEN EXPANSION FOR LINKED DUFFING LINEAR OSCILLATORS EXCITED BY COLORED NOISE

THE VAN KAMPEN EXPANSION FOR LINKED DUFFING LINEAR OSCILLATORS EXCITED BY COLORED NOISE Journal of Soun an Vibration (1996) 191(3), 397 414 THE VAN KAMPEN EXPANSION FOR LINKED DUFFING LINEAR OSCILLATORS EXCITED BY COLORED NOISE E. M. WEINSTEIN Galaxy Scientific Corporation, 2500 English Creek

More information

Computing Exact Confidence Coefficients of Simultaneous Confidence Intervals for Multinomial Proportions and their Functions

Computing Exact Confidence Coefficients of Simultaneous Confidence Intervals for Multinomial Proportions and their Functions Working Paper 2013:5 Department of Statistics Computing Exact Confience Coefficients of Simultaneous Confience Intervals for Multinomial Proportions an their Functions Shaobo Jin Working Paper 2013:5

More information

Tutorial on Maximum Likelyhood Estimation: Parametric Density Estimation

Tutorial on Maximum Likelyhood Estimation: Parametric Density Estimation Tutorial on Maximum Likelyhoo Estimation: Parametric Density Estimation Suhir B Kylasa 03/13/2014 1 Motivation Suppose one wishes to etermine just how biase an unfair coin is. Call the probability of tossing

More information

TEMPORAL AND TIME-FREQUENCY CORRELATION-BASED BLIND SOURCE SEPARATION METHODS. Yannick DEVILLE

TEMPORAL AND TIME-FREQUENCY CORRELATION-BASED BLIND SOURCE SEPARATION METHODS. Yannick DEVILLE TEMPORAL AND TIME-FREQUENCY CORRELATION-BASED BLIND SOURCE SEPARATION METHODS Yannick DEVILLE Université Paul Sabatier Laboratoire Acoustique, Métrologie, Instrumentation Bât. 3RB2, 8 Route e Narbonne,

More information

Modelling and simulation of dependence structures in nonlife insurance with Bernstein copulas

Modelling and simulation of dependence structures in nonlife insurance with Bernstein copulas Moelling an simulation of epenence structures in nonlife insurance with Bernstein copulas Prof. Dr. Dietmar Pfeifer Dept. of Mathematics, University of Olenburg an AON Benfiel, Hamburg Dr. Doreen Straßburger

More information

A note on asymptotic formulae for one-dimensional network flow problems Carlos F. Daganzo and Karen R. Smilowitz

A note on asymptotic formulae for one-dimensional network flow problems Carlos F. Daganzo and Karen R. Smilowitz A note on asymptotic formulae for one-imensional network flow problems Carlos F. Daganzo an Karen R. Smilowitz (to appear in Annals of Operations Research) Abstract This note evelops asymptotic formulae

More information

05 The Continuum Limit and the Wave Equation

05 The Continuum Limit and the Wave Equation Utah State University DigitalCommons@USU Founations of Wave Phenomena Physics, Department of 1-1-2004 05 The Continuum Limit an the Wave Equation Charles G. Torre Department of Physics, Utah State University,

More information

A Review of Multiple Try MCMC algorithms for Signal Processing

A Review of Multiple Try MCMC algorithms for Signal Processing A Review of Multiple Try MCMC algorithms for Signal Processing Luca Martino Image Processing Lab., Universitat e València (Spain) Universia Carlos III e Mari, Leganes (Spain) Abstract Many applications

More information

Final Exam Study Guide and Practice Problems Solutions

Final Exam Study Guide and Practice Problems Solutions Final Exam Stuy Guie an Practice Problems Solutions Note: These problems are just some of the types of problems that might appear on the exam. However, to fully prepare for the exam, in aition to making

More information

Entanglement is not very useful for estimating multiple phases

Entanglement is not very useful for estimating multiple phases PHYSICAL REVIEW A 70, 032310 (2004) Entanglement is not very useful for estimating multiple phases Manuel A. Ballester* Department of Mathematics, University of Utrecht, Box 80010, 3508 TA Utrecht, The

More information

2 Viktor G. Kurotschka, Rainer Schwabe. 1) In the case of a small experimental region, mathematically described by

2 Viktor G. Kurotschka, Rainer Schwabe. 1) In the case of a small experimental region, mathematically described by HE REDUION OF DESIGN PROLEMS FOR MULIVARIAE EXPERIMENS O UNIVARIAE POSSIILIIES AND HEIR LIMIAIONS Viktor G Kurotschka, Rainer Schwabe Freie Universitat erlin, Mathematisches Institut, Arnimallee 2{6, D-4

More information

Learning in Monopolies with Delayed Price Information

Learning in Monopolies with Delayed Price Information Learning in Monopolies with Delaye Price Information Akio Matsumoto y Chuo University Ferenc Sziarovszky z University of Pécs February 28, 2013 Abstract We call the intercept of the price function with

More information

arxiv: v2 [cond-mat.stat-mech] 11 Nov 2016

arxiv: v2 [cond-mat.stat-mech] 11 Nov 2016 Noname manuscript No. (will be inserte by the eitor) Scaling properties of the number of ranom sequential asorption iterations neee to generate saturate ranom packing arxiv:607.06668v2 [con-mat.stat-mech]

More information

Survey-weighted Unit-Level Small Area Estimation

Survey-weighted Unit-Level Small Area Estimation Survey-weighte Unit-Level Small Area Estimation Jan Pablo Burgar an Patricia Dörr Abstract For evience-base regional policy making, geographically ifferentiate estimates of socio-economic inicators are

More information

Semianalytical method of lines for solving elliptic partial dierential equations

Semianalytical method of lines for solving elliptic partial dierential equations Chemical Engineering Science 59 (2004) 781 788 wwwelseviercom/locate/ces Semianalytical metho of lines for solving elliptic partial ierential equations Venkat R Subramanian a;, Ralph E White b a Department

More information

A Course in Machine Learning

A Course in Machine Learning A Course in Machine Learning Hal Daumé III 12 EFFICIENT LEARNING So far, our focus has been on moels of learning an basic algorithms for those moels. We have not place much emphasis on how to learn quickly.

More information

PAijpam.eu RELATIVE HEAT LOSS REDUCTION FORMULA FOR WINDOWS WITH MULTIPLE PANES Cassandra Reed 1, Jean Michelet Jean-Michel 2

PAijpam.eu RELATIVE HEAT LOSS REDUCTION FORMULA FOR WINDOWS WITH MULTIPLE PANES Cassandra Reed 1, Jean Michelet Jean-Michel 2 International Journal of Pure an Applie Mathematics Volume 97 No. 4 2014 543-549 ISSN: 1311-8080 (printe version); ISSN: 1314-3395 (on-line version) url: http://www.ijpam.eu oi: http://x.oi.org/10.12732/ijpam.v97i4.13

More information

Parameter estimation: A new approach to weighting a priori information

Parameter estimation: A new approach to weighting a priori information Parameter estimation: A new approach to weighting a priori information J.L. Mea Department of Mathematics, Boise State University, Boise, ID 83725-555 E-mail: jmea@boisestate.eu Abstract. We propose a

More information

Wavelet-Based Parameter Estimation for Polynomial Contaminated Fractionally Differenced Processes

Wavelet-Based Parameter Estimation for Polynomial Contaminated Fractionally Differenced Processes 1 Wavelet-Base Parameter Estimation for Polynomial Contaminate Fractionally Difference Processes Peter F. Craigmile, Peter Guttorp, an Donal B. Percival. Abstract We consier the problem of estimating the

More information

A LIMIT THEOREM FOR RANDOM FIELDS WITH A SINGULARITY IN THE SPECTRUM

A LIMIT THEOREM FOR RANDOM FIELDS WITH A SINGULARITY IN THE SPECTRUM Teor Imov r. ta Matem. Statist. Theor. Probability an Math. Statist. Vip. 81, 1 No. 81, 1, Pages 147 158 S 94-911)816- Article electronically publishe on January, 11 UDC 519.1 A LIMIT THEOREM FOR RANDOM

More information

Schrödinger s equation.

Schrödinger s equation. Physics 342 Lecture 5 Schröinger s Equation Lecture 5 Physics 342 Quantum Mechanics I Wenesay, February 3r, 2010 Toay we iscuss Schröinger s equation an show that it supports the basic interpretation of

More information

Designing of Acceptance Double Sampling Plan for Life Test Based on Percentiles of Exponentiated Rayleigh Distribution

Designing of Acceptance Double Sampling Plan for Life Test Based on Percentiles of Exponentiated Rayleigh Distribution International Journal of Statistics an Systems ISSN 973-675 Volume, Number 3 (7), pp. 475-484 Research Inia Publications http://www.ripublication.com Designing of Acceptance Double Sampling Plan for Life

More information

Analytic Scaling Formulas for Crossed Laser Acceleration in Vacuum

Analytic Scaling Formulas for Crossed Laser Acceleration in Vacuum October 6, 4 ARDB Note Analytic Scaling Formulas for Crosse Laser Acceleration in Vacuum Robert J. Noble Stanfor Linear Accelerator Center, Stanfor University 575 San Hill Roa, Menlo Park, California 945

More information

SYNCHRONOUS SEQUENTIAL CIRCUITS

SYNCHRONOUS SEQUENTIAL CIRCUITS CHAPTER SYNCHRONOUS SEUENTIAL CIRCUITS Registers an counters, two very common synchronous sequential circuits, are introuce in this chapter. Register is a igital circuit for storing information. Contents

More information

Time series power spectral density. frequency-side,, vs. time-side, t

Time series power spectral density. frequency-side,, vs. time-side, t ime series power spectral ensity. requency-sie,, vs. time-sie, t t, t= 0, ±1, ±, Suppose stationary c u = cov{t+u,t} u = 0, ±1, ±, lag = 1/π Σ exp {-iu} c u perio π non-negative / : requency cycles/unit

More information

One-dimensional I test and direction vector I test with array references by induction variable

One-dimensional I test and direction vector I test with array references by induction variable Int. J. High Performance Computing an Networking, Vol. 3, No. 4, 2005 219 One-imensional I test an irection vector I test with array references by inuction variable Minyi Guo School of Computer Science

More information

Quantile function expansion using regularly varying functions

Quantile function expansion using regularly varying functions Quantile function expansion using regularly varying functions arxiv:705.09494v [math.st] 9 Aug 07 Thomas Fung a, an Eugene Seneta b a Department of Statistics, Macquarie University, NSW 09, Australia b

More information

. Using a multinomial model gives us the following equation for P d. , with respect to same length term sequences.

. Using a multinomial model gives us the following equation for P d. , with respect to same length term sequences. S 63 Lecture 8 2/2/26 Lecturer Lillian Lee Scribes Peter Babinski, Davi Lin Basic Language Moeling Approach I. Special ase of LM-base Approach a. Recap of Formulas an Terms b. Fixing θ? c. About that Multinomial

More information

Math 342 Partial Differential Equations «Viktor Grigoryan

Math 342 Partial Differential Equations «Viktor Grigoryan Math 342 Partial Differential Equations «Viktor Grigoryan 6 Wave equation: solution In this lecture we will solve the wave equation on the entire real line x R. This correspons to a string of infinite

More information

Calculus and optimization

Calculus and optimization Calculus an optimization These notes essentially correspon to mathematical appenix 2 in the text. 1 Functions of a single variable Now that we have e ne functions we turn our attention to calculus. A function

More information

β ˆ j, and the SD path uses the local gradient

β ˆ j, and the SD path uses the local gradient Proceeings of the 00 Winter Simulation Conference E. Yücesan, C.-H. Chen, J. L. Snowon, an J. M. Charnes, es. RESPONSE SURFACE METHODOLOGY REVISITED Ebru Angün Jack P.C. Kleijnen Department of Information

More information

Image Denoising Using Spatial Adaptive Thresholding

Image Denoising Using Spatial Adaptive Thresholding International Journal of Engineering Technology, Management an Applie Sciences Image Denoising Using Spatial Aaptive Thresholing Raneesh Mishra M. Tech Stuent, Department of Electronics & Communication,

More information

Heteroscedasticityinstochastic frontier models: A Monte Carlo Analysis

Heteroscedasticityinstochastic frontier models: A Monte Carlo Analysis Heterosceasticityinstochastic frontier moels: A Monte Carlo Analysis by C. Guermat University of Exeter, Exeter EX4 4RJ, UK an K. Hari City University, Northampton Square, Lonon EC1V 0HB, UK First version:

More information

Linköping University Electronic Press

Linköping University Electronic Press Linköping University Electronic Press Report On the Multivariate t Distribution Michael Roth LiTH-ISY-R, 400-390, No. 3059 Available at: Linköping University Electronic Press http://urn.kb.se/resolve?urn=urn:nbn:se:liu:iva-9686

More information

Energy behaviour of the Boris method for charged-particle dynamics

Energy behaviour of the Boris method for charged-particle dynamics Version of 25 April 218 Energy behaviour of the Boris metho for charge-particle ynamics Ernst Hairer 1, Christian Lubich 2 Abstract The Boris algorithm is a wiely use numerical integrator for the motion

More information

Optimal Measurement and Control in Quantum Dynamical Systems.

Optimal Measurement and Control in Quantum Dynamical Systems. Optimal Measurement an Control in Quantum Dynamical Systems. V P Belavin Institute of Physics, Copernicus University, Polan. (On leave of absence from MIEM, Moscow, USSR) Preprint No 411, Torun, February

More information

d dx But have you ever seen a derivation of these results? We ll prove the first result below. cos h 1

d dx But have you ever seen a derivation of these results? We ll prove the first result below. cos h 1 Lecture 5 Some ifferentiation rules Trigonometric functions (Relevant section from Stewart, Seventh Eition: Section 3.3) You all know that sin = cos cos = sin. () But have you ever seen a erivation of

More information

Time Series: Theory and Methods

Time Series: Theory and Methods Peter J. Brockwell Richard A. Davis Time Series: Theory and Methods Second Edition With 124 Illustrations Springer Contents Preface to the Second Edition Preface to the First Edition vn ix CHAPTER 1 Stationary

More information

STATISTICAL LIKELIHOOD REPRESENTATIONS OF PRIOR KNOWLEDGE IN MACHINE LEARNING

STATISTICAL LIKELIHOOD REPRESENTATIONS OF PRIOR KNOWLEDGE IN MACHINE LEARNING STATISTICAL LIKELIHOOD REPRESENTATIONS OF PRIOR KNOWLEDGE IN MACHINE LEARNING Mark A. Kon Department of Mathematics an Statistics Boston University Boston, MA 02215 email: mkon@bu.eu Anrzej Przybyszewski

More information

The effect of nonvertical shear on turbulence in a stably stratified medium

The effect of nonvertical shear on turbulence in a stably stratified medium The effect of nonvertical shear on turbulence in a stably stratifie meium Frank G. Jacobitz an Sutanu Sarkar Citation: Physics of Fluis (1994-present) 10, 1158 (1998); oi: 10.1063/1.869640 View online:

More information

'HVLJQ &RQVLGHUDWLRQ LQ 0DWHULDO 6HOHFWLRQ 'HVLJQ 6HQVLWLYLW\,1752'8&7,21

'HVLJQ &RQVLGHUDWLRQ LQ 0DWHULDO 6HOHFWLRQ 'HVLJQ 6HQVLWLYLW\,1752'8&7,21 Large amping in a structural material may be either esirable or unesirable, epening on the engineering application at han. For example, amping is a esirable property to the esigner concerne with limiting

More information

How to Minimize Maximum Regret in Repeated Decision-Making

How to Minimize Maximum Regret in Repeated Decision-Making How to Minimize Maximum Regret in Repeate Decision-Making Karl H. Schlag July 3 2003 Economics Department, European University Institute, Via ella Piazzuola 43, 033 Florence, Italy, Tel: 0039-0-4689, email:

More information

A. Exclusive KL View of the MLE

A. Exclusive KL View of the MLE A. Exclusive KL View of the MLE Lets assume a change-of-variable moel p Z z on the ranom variable Z R m, such as the one use in Dinh et al. 2017: z 0 p 0 z 0 an z = ψz 0, where ψ is an invertible function

More information

Similarity Measures for Categorical Data A Comparative Study. Technical Report

Similarity Measures for Categorical Data A Comparative Study. Technical Report Similarity Measures for Categorical Data A Comparative Stuy Technical Report Department of Computer Science an Engineering University of Minnesota 4-92 EECS Builing 200 Union Street SE Minneapolis, MN

More information

Lecture Introduction. 2 Examples of Measure Concentration. 3 The Johnson-Lindenstrauss Lemma. CS-621 Theory Gems November 28, 2012

Lecture Introduction. 2 Examples of Measure Concentration. 3 The Johnson-Lindenstrauss Lemma. CS-621 Theory Gems November 28, 2012 CS-6 Theory Gems November 8, 0 Lecture Lecturer: Alesaner Mąry Scribes: Alhussein Fawzi, Dorina Thanou Introuction Toay, we will briefly iscuss an important technique in probability theory measure concentration

More information

19 Eigenvalues, Eigenvectors, Ordinary Differential Equations, and Control

19 Eigenvalues, Eigenvectors, Ordinary Differential Equations, and Control 19 Eigenvalues, Eigenvectors, Orinary Differential Equations, an Control This section introuces eigenvalues an eigenvectors of a matrix, an iscusses the role of the eigenvalues in etermining the behavior

More information

A simple tranformation of copulas

A simple tranformation of copulas A simple tranformation of copulas V. Durrleman, A. Nikeghbali & T. Roncalli Groupe e Recherche Opérationnelle Créit Lyonnais France July 31, 2000 Abstract We stuy how copulas properties are moifie after

More information

Role of parameters in the stochastic dynamics of a stick-slip oscillator

Role of parameters in the stochastic dynamics of a stick-slip oscillator Proceeing Series of the Brazilian Society of Applie an Computational Mathematics, v. 6, n. 1, 218. Trabalho apresentao no XXXVII CNMAC, S.J. os Campos - SP, 217. Proceeing Series of the Brazilian Society

More information

FLUCTUATIONS IN THE NUMBER OF POINTS ON SMOOTH PLANE CURVES OVER FINITE FIELDS. 1. Introduction

FLUCTUATIONS IN THE NUMBER OF POINTS ON SMOOTH PLANE CURVES OVER FINITE FIELDS. 1. Introduction FLUCTUATIONS IN THE NUMBER OF POINTS ON SMOOTH PLANE CURVES OVER FINITE FIELDS ALINA BUCUR, CHANTAL DAVID, BROOKE FEIGON, MATILDE LALÍN 1 Introuction In this note, we stuy the fluctuations in the number

More information

Conservation laws a simple application to the telegraph equation

Conservation laws a simple application to the telegraph equation J Comput Electron 2008 7: 47 51 DOI 10.1007/s10825-008-0250-2 Conservation laws a simple application to the telegraph equation Uwe Norbrock Reinhol Kienzler Publishe online: 1 May 2008 Springer Scienceusiness

More information

Time Series Analysis. Correlated Errors in the Parameters Estimation of the ARFIMA Model: A Simulated Study

Time Series Analysis. Correlated Errors in the Parameters Estimation of the ARFIMA Model: A Simulated Study Communications in Statistics Simulation and Computation, 35: 789 802, 2006 Copyright Taylor & Francis Group, LLC ISSN: 0361-0918 print/1532-4141 online DOI: 10.1080/03610910600716928 Time Series Analysis

More information

DAMTP 000/NA04 On the semi-norm of raial basis function interpolants H.-M. Gutmann Abstract: Raial basis function interpolation has attracte a lot of

DAMTP 000/NA04 On the semi-norm of raial basis function interpolants H.-M. Gutmann Abstract: Raial basis function interpolation has attracte a lot of UNIVERSITY OF CAMBRIDGE Numerical Analysis Reports On the semi-norm of raial basis function interpolants H.-M. Gutmann DAMTP 000/NA04 May, 000 Department of Applie Mathematics an Theoretical Physics Silver

More information

Acute sets in Euclidean spaces

Acute sets in Euclidean spaces Acute sets in Eucliean spaces Viktor Harangi April, 011 Abstract A finite set H in R is calle an acute set if any angle etermine by three points of H is acute. We examine the maximal carinality α() of

More information

Polynomial Inclusion Functions

Polynomial Inclusion Functions Polynomial Inclusion Functions E. e Weert, E. van Kampen, Q. P. Chu, an J. A. Muler Delft University of Technology, Faculty of Aerospace Engineering, Control an Simulation Division E.eWeert@TUDelft.nl

More information

Hyperbolic Systems of Equations Posed on Erroneous Curved Domains

Hyperbolic Systems of Equations Posed on Erroneous Curved Domains Hyperbolic Systems of Equations Pose on Erroneous Curve Domains Jan Norström a, Samira Nikkar b a Department of Mathematics, Computational Mathematics, Linköping University, SE-58 83 Linköping, Sween (

More information

Math 115 Section 018 Course Note

Math 115 Section 018 Course Note Course Note 1 General Functions Definition 1.1. A function is a rule that takes certain numbers as inputs an assigns to each a efinite output number. The set of all input numbers is calle the omain of

More information

CONTROL CHARTS FOR VARIABLES

CONTROL CHARTS FOR VARIABLES UNIT CONTOL CHATS FO VAIABLES Structure.1 Introuction Objectives. Control Chart Technique.3 Control Charts for Variables.4 Control Chart for Mean(-Chart).5 ange Chart (-Chart).6 Stanar Deviation Chart

More information

The Impact of Collusion on the Price of Anarchy in Nonatomic and Discrete Network Games

The Impact of Collusion on the Price of Anarchy in Nonatomic and Discrete Network Games The Impact of Collusion on the Price of Anarchy in Nonatomic an Discrete Network Games Tobias Harks Institute of Mathematics, Technical University Berlin, Germany harks@math.tu-berlin.e Abstract. Hayrapetyan,

More information

New Statistical Test for Quality Control in High Dimension Data Set

New Statistical Test for Quality Control in High Dimension Data Set International Journal of Applie Engineering Research ISSN 973-456 Volume, Number 6 (7) pp. 64-649 New Statistical Test for Quality Control in High Dimension Data Set Shamshuritawati Sharif, Suzilah Ismail

More information

inflow outflow Part I. Regular tasks for MAE598/494 Task 1

inflow outflow Part I. Regular tasks for MAE598/494 Task 1 MAE 494/598, Fall 2016 Project #1 (Regular tasks = 20 points) Har copy of report is ue at the start of class on the ue ate. The rules on collaboration will be release separately. Please always follow the

More information

Prof. Dr. Ibraheem Nasser electric_charhe 9/22/2017 ELECTRIC CHARGE

Prof. Dr. Ibraheem Nasser electric_charhe 9/22/2017 ELECTRIC CHARGE ELECTRIC CHARGE Introuction: Orinary matter consists of atoms. Each atom consists of a nucleus, consisting of protons an neutrons, surroune by a number of electrons. In electricity, the electric charge

More information

A Randomized Approximate Nearest Neighbors Algorithm - a short version

A Randomized Approximate Nearest Neighbors Algorithm - a short version We present a ranomize algorithm for the approximate nearest neighbor problem in - imensional Eucliean space. Given N points {x } in R, the algorithm attempts to fin k nearest neighbors for each of x, where

More information

A PAC-Bayesian Approach to Spectrally-Normalized Margin Bounds for Neural Networks

A PAC-Bayesian Approach to Spectrally-Normalized Margin Bounds for Neural Networks A PAC-Bayesian Approach to Spectrally-Normalize Margin Bouns for Neural Networks Behnam Neyshabur, Srinah Bhojanapalli, Davi McAllester, Nathan Srebro Toyota Technological Institute at Chicago {bneyshabur,

More information

A Simple Model for the Calculation of Plasma Impedance in Atmospheric Radio Frequency Discharges

A Simple Model for the Calculation of Plasma Impedance in Atmospheric Radio Frequency Discharges Plasma Science an Technology, Vol.16, No.1, Oct. 214 A Simple Moel for the Calculation of Plasma Impeance in Atmospheric Raio Frequency Discharges GE Lei ( ) an ZHANG Yuantao ( ) Shanong Provincial Key

More information

EVALUATING HIGHER DERIVATIVE TENSORS BY FORWARD PROPAGATION OF UNIVARIATE TAYLOR SERIES

EVALUATING HIGHER DERIVATIVE TENSORS BY FORWARD PROPAGATION OF UNIVARIATE TAYLOR SERIES MATHEMATICS OF COMPUTATION Volume 69, Number 231, Pages 1117 1130 S 0025-5718(00)01120-0 Article electronically publishe on February 17, 2000 EVALUATING HIGHER DERIVATIVE TENSORS BY FORWARD PROPAGATION

More information

Modelling long-term heart rate variability: an ARFIMA approach

Modelling long-term heart rate variability: an ARFIMA approach Biome Tech 006; 51:15 19 006 by Walter e Gruyter Berlin New York. DOI 10.1515/BMT.006.040 Moelling long-term heart rate variability: an ARFIMA approach Argentina S. Leite 1 3, *, Ana Paula Rocha 1,3, M.

More information

The derivative of a function f(x) is another function, defined in terms of a limiting expression: f(x + δx) f(x)

The derivative of a function f(x) is another function, defined in terms of a limiting expression: f(x + δx) f(x) Y. D. Chong (2016) MH2801: Complex Methos for the Sciences 1. Derivatives The erivative of a function f(x) is another function, efine in terms of a limiting expression: f (x) f (x) lim x δx 0 f(x + δx)

More information

Optimization of Geometries by Energy Minimization

Optimization of Geometries by Energy Minimization Optimization of Geometries by Energy Minimization by Tracy P. Hamilton Department of Chemistry University of Alabama at Birmingham Birmingham, AL 3594-140 hamilton@uab.eu Copyright Tracy P. Hamilton, 1997.

More information

A Path Planning Method Using Cubic Spiral with Curvature Constraint

A Path Planning Method Using Cubic Spiral with Curvature Constraint A Path Planning Metho Using Cubic Spiral with Curvature Constraint Tzu-Chen Liang an Jing-Sin Liu Institute of Information Science 0, Acaemia Sinica, Nankang, Taipei 5, Taiwan, R.O.C., Email: hartree@iis.sinica.eu.tw

More information

Topic 7: Convergence of Random Variables

Topic 7: Convergence of Random Variables Topic 7: Convergence of Ranom Variables Course 003, 2016 Page 0 The Inference Problem So far, our starting point has been a given probability space (S, F, P). We now look at how to generate information

More information

The effect of dissipation on solutions of the complex KdV equation

The effect of dissipation on solutions of the complex KdV equation Mathematics an Computers in Simulation 69 (25) 589 599 The effect of issipation on solutions of the complex KV equation Jiahong Wu a,, Juan-Ming Yuan a,b a Department of Mathematics, Oklahoma State University,

More information