Optimal Jackknife for Discrete Time and Continuous Time Unit Root Models

Size: px
Start display at page:

Download "Optimal Jackknife for Discrete Time and Continuous Time Unit Root Models"

Transcription

1 Optial Jackknife for Discrete Tie and Continuous Tie Unit Root Models Ye Chen and Jun Yu Singapore Manageent University January 6, Abstract Maxiu likelihood estiation of the persistence paraeter in the discrete tie unit root odel is known for suffering fro a downward bias. The bias is ore pronounced in the continuous tie unit root odel. Recently Chabers and Kyriacou introduced a new jackknife ethod to reove the first order bias in the estiator of the persistence paraeter in a discrete tie unit root odel. This paper proposes an iproved jackknife estiator of the persistence paraeter that works for both the discrete tie unit root odel and the continuous tie unit root odel. The proposed jackknife estiator is optial in the sense that it iniizes the variance. Siulations highlight the perforance of the proposed ethod in both contexts. They show that our optial jackknife reduces the variance of the jackknife ethod of Chabers and Kyriacou by at least % in both cases. Keywords: Bias reduction, Variance reduction, Vasicek odel, Long-span Asyptotics, Autoregression JEL classification: C, C5 Introduction As a subsapling ethod, the jackknife estiator was first proposed by Quenouille 99 for reducing the bias in the estiation of the serial correlation coeffi cient. Tukey 958 used the jackknife ethod to estiate the variance. Miller 97 reviewed the literature of the jackknife, ephasizing its properties in bias reduction and interval estiation. More recently, the jackknife ethod has been found useful Ye Chen, School of Econoics and Si Kee Boon Institute for Financial Econoics, Singapore Manageent University, 9 Staford Road, Singapore, 7893; Eail: chen.ye.9@phdecons.su.edu.sg. Jun Yu, Si Kee Boon Institute for Financial Econoics, School of Econoics and Lee Kong Chian School of Business, Singapore Manageent University, 9 Staford Road, Singapore Eail: yujun@su.edu.sg. We acknowledge coents fro Peter Phillips and Denis Leung.

2 in econoetrics. Hahn and Newey 3 deonstrated the use of the jackknife to reduce the bias arising fro the incidental paraeters proble in the dynaic panel odel with fixed effects. Phillips and Yu 5, PY hereafter introduced a jackknife ethod based on non-overlapping subsaples, and showed that it can be applied to the coeffi cients in continuous tie odels as well as the asset prices directly. Chiquoine and Hjalarsson 9 provided the evidence that the jackknife can be successfully applied to stock return predictability regressions. In the tie series context, the jackknife ethod proposed by PY is very easy to ipleent and also quite effective in reducing the bias. In effect, this jackknife estiator reoves the first order bias in the original estiator by taking a linear cobination of the full saple estiator and a set of subsaple estiators. The validity for reoving the first order bias can be explained by the Nagar approxiation of the oents of the original estiator, in ters of the oents of the estiator s Taylor expansion as a polynoial of the saple oents of the data. In particular, under regular conditions, the first order bias is a reciprocal function of the saple size with a coon coeffi cient. Consequently, the weights used for the subsaples are the sae and are proportional to the weight used for the full saple. For stationary autoregressive odels, Chabers obtained the liit distribution of the jackknife estiator of PY and exained its perforance relative to alternative jackknifing procedures. However, in the context of a discrete tie unit root odel, Chabers and Kyriacou, CK hereafter pointed out that the jackknife estiator of PY cannot copletely reove the first order bias. This is because the first order bias of the least squares LS or axiu likelihood ML estiator of the autoregressive coeffi cient has distinctive functional fors in different subsaples. This distinctiveness is anifest in the fact that the liit distribution of the LS/ML estiator depends on the initial condition. A revised jackknife estiator with new weights was proposed in CK. Unlike the PY estiator, the weight used in CK for the full saple is no longer proportional to those for the subsaples. However, the weights used for the subsaples are required to be identical to each other in CK. CK showed that the odified jackknife estiate perfors better than the PY estiator for bias reduction. While the jackknife ethod of CK reduces the bias of the original estiator, it always increases the variance. This is not surprising because subsaple estiators are used in the construction of the jackknife. As a result, the trade-off between the bias reduction and the increase in variance is iportant. Siulation results in CK suggest that their jackknife ethod soeties gains enough in bias reduction without a significant increase in variance, so that there ay be an overall gain in root ean square error RMSE for the discrete tie unit root odel. In this paper, we propose an iproved jackknife estiator for unit root odels. Our estiator is optial in the sense that it not only reoves the first order bias, but also iniizes the variance, and hence, has better finite saple properties than the CK estiator. Like the estiators of PY and CK,

3 our optial jackknife estiator is a linear cobination of the full saple estiator and the subsaple estiators. However, we do not require the weight for the full saple to be proportional to those for the subsaple, nor do we require the weights for the subsaple be the sae. Indeed, all the weights are obtained by iniizing the variance under the condition that the first order bias is reoved. The calculation of the weights involves obtaining the covariances between the full saple estiator and the subsaple estiators, and the covariances between the subsaple estiators. To do so, we eploy the joint oent generating function MGF of the liit distributions of these estiators. The idea is applied to both the discrete tie and the continuous tie unit root odels. Optial weights are derived and the finite saple perforance of the new estiator is exained for both odels. It is found that the optial jackknife estiator offers over % reduction in variance over the CK estiator without coproising bias reduction in both cases. The paper is organized as follows. Section derives the general for of the optial jackknife estiator in the discrete tie unit root odel. Section 3 extends the results to the continuous tie unit root odel where the bias of the persistent paraeter is known to be ore serious. Section presents the Monte Carlo siulation results. Section 5 concludes. Appendix collects all the proofs. Throughout the paper, we adopt the following notations. The full saple size is denoted by n, the tie span of data by T, the paraeter of interest by θ or β or κ, and the true value of it by θ or β or κ. The signal is used to indicate the convergence of the associated probability easures as n or T goes to infinity while denotes the equality in distribution, and the nuber of subsaples. θ j eans the LS/ML estiator of θ fro the jth subsaple of saple length l i.e., l = n, θ P Y is the jackknife estiator of θ proposed by PY, θ CK is the jackknife estiator proposed CY by CK, and θ is the jackknife estiator proposed in the present paper. Following CK, we define Z = W dw/ W, Z j = j/ j / W dw/ j/ j / W, µ = EZ and µ j = EZ j, where W is a standard Brownian otion. Optial Jackknife for the Discrete Tie Unit Root Model In this section, we first briefly review the literature, focusing especially on the approach of CK. We then introduce our optial jackknife estiator in the context of the discrete tie unit root odel.. A Literature Review Considering a siple unit root odel with initial value y = O p : y t = y t + ε t, ε t iid, σ ε, t =,,..., n.. 3

4 The LS estiator of the autoregressive AR paraeter of, say, β, is β = n t= y t y t / n t= y t. When ε t is norally distributed, β is also the ML estiator of β, conditional on y. The liit distribution of β was obtained by Phillips 987a using the functional central liit theory and the continuous apping theore: n β W dw W.. The liit distribution is skewed for two reasons. First, W dw is not centered around zero. Second, W dw and W are dependent. This skewness gives rise to the first order bias in the LS/ML estiator, even asyptotically. Under the assuption that the initial value y is, Phillips 987a obtained the asyptotic expansion of the liit distribution of n β : n β = W dw W / nξ W + O pn,.3 where ξ N,, and is independent of W. Taking the expectation of.3 and noting that µ = EZ =.78, we have E β =.78 n + On.. This result is in contrast to the Kendall bias forula for the stationary AR odel i.e. β < : when E β β = β n + On..5 Following the original work of Quenouille, PY 5 utilized the subsaple estiators of β to achieve bias reduction with the following linear cobination: β P Y = β j= β j,.6 where β is the LS/ML estiator of β based on the full saple, i.e., y,..., y n ; βj is the LS/ML estiator of β based on the jth subsaple, i.e., y j l+,..., y jl. There are two iportant features in this estiator. First, the weight given to the full saple estiator is proportional to i.e. ties as large as that assigned to the average of the subsaple estiators. Second, the weights are the sae for all the subsaple estiators. The validity of this jackknife ethod can be explained by a general result based on the Nagar expansion: E β = β + b n + O n,.7

5 which is assued to hold for soe constant b. Two observations can be ade fro this general result. First, to the first order, the bias is a reciprocal function of the saple size. This explains why in the PY ethod, the weight assigned to the full saple estiator is ties as big as that assigned to the subsaple estiators. Second, to the first order, the bias is the sae across different subsaples. This explains why in the PY ethod, the weights are the sae for all the subsaple estiators. It is Y easy to show that the jackknife estiator reoves the first order bias, i.e., E βp = β + O n, when.7 holds true. Particularly effective bias reduction can be achieved by choosing =. In this case, this estiator becoes: β P Y = β β + β..8 Both PY and Chabers have reported evidence to support this ethod for the purpose of bias reduction in different contexts. It should be pointed out that any jackknife procedures have been proposed, including the delete- jackknife of Tukey 958, the delete-d jackknife of Wu 986, and the oving-block jackknife of Chabers. The Nagar approxiation is a general result and ay be verified by Sargan s 976 theore. Given the ild conditions under which Sargan s 976 theore hold, it is perhaps rather surprising that the jackknife ethod fails to reove the first order bias in a unit root odel. This failure was first docuented in CK. The basic arguent of CK is that in.7, b is not a constant any ore in the unit root odel. Instead, it depends on the initial condition. As the initial condition varies across different subsaples, the jackknife cannot eliinate the first order bias ter. To be ore specific, the liit distribution of the noralized subsaples estiator, l β j, is j/ j / W dw/ j/. j / W j/ CK showed that µ j = E j / W dw/ j/ j / W depends on j. To eliinate the first order asyptotic bias, CK proposed the following odified jackknife estiator: where β CK = b CK β j= δ CK β j,.9 b CK j= = µ j j= µ j µ, µ δck =.. j= µ j µ When µ = = µ = µ, b CK = / = b P Y, and δ CK = /, the CK estiator is the sae as the PY estiator. Under odel., CK showed that µ = µ =.783, µ =.389, µ 3 =.9399, µ =.833, etc. That is, the bias becoes saller and saller as we go deeper and deeper into subsapling. Substituting these expected values into the forulae., we can calculate the weights. Table reports the weights when =. We also report the weights of PY for coparison. As µ is closer to zero than µ and µ, a larger weight is assigned to the full saple 5

6 Table : Weights assigned to the full saple and subsaple estiators for alternative jackknife ethods b P Y b CK b CY a CY, a CY, δ P Y δ CK estiator, copared to the PY estiator. The dependence of the bias on the initial condition is a very interesting finding which is explained by a recent result obtained in Phillips and Magdalinos 9, PM hereafter. In PM, the initial condition, y, is assued to be: with y n = k n j= ε j,. k n /n τ, + ].. When τ =, y n is said to be a recent past initialization, which obviously includes y = as a special case. When τ =, +, y n is said to be a distant past initialization. When τ =, y n is said to be an infinite past initialization. PM showed that when τ, +, the liit distribution of the LS/ML estiator of β under the unit root odel is given by: n β W τ + dw W +, τ where W τ + t = W t + τw t, and W is another standard Brownian otion but independent of W. If τ =, the liit distribution is the sae as in.. When τ = +, the liit distribution of the LS/ML estiator of β under the unit root odel is given by: kn n β C, where C stands for a Cauchy variate. The dependence of the liit distribution and hence its first oent on the initial condition are clearly seen fro the above results. The larger the value of τ, the greater is the iportance of the initial condition, and the bigger is the share of τw t in W τ + t. For Model., τ = applied for the first saple and τ = jl/n, for the j + th subsaple. That is, for j =,,...,, we have τ =, /,..., /. As we go deeper and deeper into subsapling, the bigger and bigger is the influence of the initial condition on the liit distribution. Not surprisingly, all the oents, including the ean and the variance, are different for different subsaples. 6

7 . Optial Jackknife It is known that the jackknife estiator of PY increases the variance, when copared to the LS/ML estiator. The sae feature is seen in the estiator of CK. In this paper, we introduce a new jackknife estiator, which can reove the first order bias as well as iniize the variance for a given. The reason why it is possible to reduce the variance of the estiator of CK is that it is not necessary to assign an identical weight to different subsaples, given that the biases of these estiators are different. To fix the idea, the new jackknife estiator is defined as: β CY = b CY β j= a CY j, β j, where b CY and {a CY j, } j= are the weights assigned to the full saple estiator and the subsaple estiators, respectively. Unlike the jackknife estiators of PY and CK, we do not require the weights for different subsaples to be the sae. iniize the variance of the new estiator, i.e.: subject to two constraints: b CY in,{acy j, } j= Our objective is to select the weights b CY, {a CY j, } j=, to V ar b CY = b CY j= b CY µ = β a CY j, β j j=,.3 a CY j, +,. j= a CY j,µ j,.5 where µ = µ. These two constraints are used to ensure the first order bias of the LS/ML estiator is fully reoved. Fro the two constraints, we can express b CY and a CY, as functions of a CY,,..., a CY,: Substituting b CY respect to a CY j,, we get: b CY = a CY µ µ, µ + + µ µ acy, µ +, a CY, = a CY µ µ, µ + + µ µ acy, µ +. and a CY, into the objective function.3, and then taking the derivative with 7

8 = b CY µ µ j +a CY, + for j =,,. i= ] µ Cov β, β Cov β, β j µ V ar β µ µ ] j µ Cov β, β + Cov β, β j + µ µ i µ Cov β, β i + µ µ ] i µ Cov β, β i + Cov β i, β j, µ V ar β µ µ j µ µ j a CY i, a CY,,..., a CY, and hence b CY These first order conditions can be used to obtain analytical expressions for and a CY,, all as functions of µ, µ,..., µ, the variances and the covariances of the full saple and subsaple estiators. To eliinate the first order bias, one ust first obtain µ, µ,..., µ, as CK did. To iniize the variance of the new estiator, one ust also calculate the exact variances and covariances of the finite saple distributions. However, it is known in the literature that the exact oents are analytically very diffi cult to obtain in dynaic odels. To siplify the derivations, we approxiate the oents of the finite saple distributions by those of the liit distributions. We shall check the quality of these approxiations in the siulation studies. While the techniques proposed in White 96 and in CK can be cobined to copute the variances, additional effort is needed to copute the covariances. A technical contribution of the present paper is to show how to copute these covariances. We now illustrate the optial jackknife ethod in a special case where =, that is, we split the full saple into two non-overlapping subsaples. The first subsaple is ade with the observations fro y to y l, with l = n/ and the initial condition, and the reainder belongs to the second subsaple with the initial condition y l. Under the two non-overlapping subsaple schee, we have: β CY = b CY β a CY β + a CY β. The objective function and the constraints are: in b CY,a CY,a CY s.t. b CY V ar β + j= b CY = a CY + a CY +, a CY j b CY µ = a CY µ + a CY µ. V ar βj b CY j= a CY j Cov β, β j + a CY a CY Cov β, β, Fro the two constraints, we express b CY and a CY in ters of a CY, that is: 8

9 b CY = acy µ µ +, µ.6 a CY = acy µ µ +. µ.7 Fro CK, we know that µ = µ =.783, and µ =.389, and hence, we have b CY =.7a CY + and a CY =.779a CY +. The first order condition with respect to a CY gives:.888v ar β.5558v ar β =.336Cov β, β Cov β, β + Cov β, β.9v ar β +.55V ar β + V ar β +.86Cov β, β.888cov β, β.6cov β, β. a CY.8 It is straightforward to check that this is the global iniizer as the objective function is quadratic and convex. To calculate the weights, it is iperative to obtain V ar β, V ar β, V ar β, Cov β, β, Cov β, β, and Cov β, β. Instead of obtaining the variances and the covariances fro the finite saple distributions, we will calculate the fro the corresponding asyptotic distributions. First, the variances can be coputed by cobining the techniques of White 96 and CK. Note that: n V ar β = n E β β ] E = n E W dw/ W + + n = E n + E W dw/ W W dw W µ + o, n E W dw/ W + n + on E W dw/ ] W n + on l V ar β j = E j/ j / W dw j/ j / W µ j + o, j =,. We need to copute the first ter on the right hand side of these two equations. Let Na, b = b a W dw, Da, b = b a W a < b, and M a,b θ, θ denote the joint oent generating function MGF of Na, b and Da, b. White 96 gave the forula for calculating the 9

10 second oent of Na, b/da, b as: E Na, b = Da, b w M a,b θ, θ θ= dθ dw. Following Phillips 987b, CK obtained the expression for M a,b θ, θ : θ M a,b θ, θ = exp θ ] { b a λb a] λ where λ = θ. The following proposition calculates the variances. θ + aθ λ ] } / λb a], Proposition. The second derivative of M a,b with respect to θ, evaluated at θ =, is given by.9 b a H / + H 3/ b ah H ] + 3 H 5/ H := va, b,. where H, H, H, H are given in the Appendix. When a, b] =, ], v, = / λ λ 3/ λ λ + 3 λ 5/ λ λ. This leads to the approxiate variance for the full saple estiator and the first subsaple estiator in the discrete tie unit root odel: n V ar β = l V ar β =.3 + On.. When a, b] = /, ], v/, = λ λ + λ ] / λ λ λ + λ λ ] 5/ λ λ. + λ λ ] 3/ λ This leads to the approxiate variance for the second subsaple estiator in the discrete tie unit root odel: l V ar β = On.. Reark. The variance of the full saple estiator, β, noralized by n, is.3. Interestingly, it is the sae as that of the first subsaple estiator, noralized by l. This equality arises because the full saple has the sae initial value as the first subsaple. Reark.3 The variance of the first subsaple estiator is about twice as large as that of the second subsaple estiator. This diff erence is due to the distinctive initial conditions and ay be ade clearer

11 Table : Variances of subsaple estiators j Variance noralized by l by exaining the liit distribution of l β j, l β j j/ j / W dw j/ j / W..3 In.3, the nuerator plays an insignificant role in the variance calculation since it can be rewritten as a difference between two chi-square variates. The iportant part is in the denoinator, which is a partial su of chi-square variates. For the first subsaple estiator, the denoinator starts with the square of zero. However, the denoinator of the second subsaple begins with the square of a rando variable whose agnitude is O p n. This rando initialization of the second subsaple increases the agnitude of the denoinator and, hence, tends to reduce the absolute value of the ratio. To confir our explanation, we carried out a siple Monte Carlo study, in which data were siulated fro the Brownian otion with the saple length being set at 5, and the nuber of replications set at,. The ean of / W is.3, whereas the ean of / W is.369. Not only is the ean affected, but also the variance and the entire distribution. In this case, the two variances are. and.69, respectively. Reark. Table lists the variances of all the subsaple estiators when =. It can be seen that the variance of the subsaple estiator decreases as j increases. The largest diff erence occurs between j = and j =. If is allowed to go to infinity, the liit distribution of the jackknife estiator will be the sae as that of the LS estiator, as pointed out by CK.

12 Second, to calculate the covariances, we note that: n Cov β, β j = E W dw W j/ j / W dw j/ j / W µµ j + On, j =,, / n Cov β, β W dw = E / W dw / W / W µ µ + On. It is required to obtain the expected value of the cross product of rando variables. As for the variances, we also calculate these expected values fro the joint MGF given in the following lea. Lea.5 Let M a,b,c,d θ, θ, ϕ, ϕ denote the MGF of Na, b, Nc, d, Da, b and Dc, d with a < b and c < d. Then the expectation of Na,b Nc,d Da,b Dc,d E Na, b Nc, d = Da, b Dc, d is given by: M a,b,c,d θ, θ, ϕ, ϕ θ ϕ θ=,ϕ = dθ dϕ.. The following proposition obtains the expression for the MGF of Na, b, Nc, d, Da, b and Dc, d, and the covariances when =. Proposition.6 The MGF M,,,/ θ, θ, ϕ, ϕ is given by { λ,,,/ exp θ ϕ ; θ λ,,,/ λ,,,/ the MGF M,,/, θ, θ, ϕ, ϕ is given by: { λ,,/, exp θ ϕ ; θ + ϕ λ,,/, λ,,/, and the MGF M,/,/, θ, θ, ϕ, ϕ is given by: { λ,/,/, exp θ ϕ, ϕ λ,/,/, λ,/,/, ] ] ] η,,,/ η,,/, π,,,/ η,,,/ π,,/, η,,/, η,/,/, η,,,/ η,,/, π,/,/, η,/,/, ]} / ]} / η,/,/, where λ a,b,cd, η a,b,c,d and π a,b,c,d are given in the Appendix. These MGFs, together with forula., ]} /

13 Table 3: Approxiate values for the variances and covariances for the full saple and subsaples when = Variance n V ar β l V ar β l V ar β Covariance n Cov β, β n Cov β, β n Covβ, β lead to the approxiate covariances between the full saple estiator and the two subsaple estiators: n Cov β, β = On ;.5 n Cov β, β = On ;.6 n Cov β, β =. + On..7 Reark.7 There are several interesting findings in Proposition.6. First, the covariances between the full saple estiator and the second subsaple estiator are siilar to, but slightly larger than that between the full saple estiator and the first subsaple estiator, although the variance of the second subsaple estiator is saller. This is because the correlation between the full saple estiator and the second subsaple estiator is larger due to the increased order of agnitude of the initial condition. Second, these two covariances are uch larger than the covariance between the two subsaple estiators. This is not surprising as the data used in the two subsaples estiators do not overlap. Reark.8 Table 3 suarizes the approxiate values of the variances and covariances when =. Theore.9 The optial weights for the jackknife estiator of β in the discrete tie unit root odel when = are b CY =.839, a CY =.677, and a CY =.69. Hence, the optial jackknife estiator is β CY JK =.839 β.677 β +.69 β..8 Reark. The optial jackknife estiator copares interestingly to the estiator of CK: β CK JK =.565 β.785 β β..9 3

14 Relative to the CK estiator, our estiator gives a larger weight to the full saple estiator, and ore so, to the second subsaple estiator. The weight for the second subsaple is nearly twice as uch as that for the first subsaple. Since the variance of the second subsaple estiator is nearly half as that of the first saple estiator, not surprisingly, the variance of our estiator is saller than that of CK s. 3 Optial Jackknife for the Continuous Tie Unit Root Model In the section, we extend the result to a continuous tie unit root odel. In the continuous tie stationary odels, it is well known that the estiation bias of the ean reversion paraeter depends on the span of the data, but not the nuber of observations. In epirically realistic situations, the tie span easured in nuber of years is often sall. Consequently, the bias can be uch ore substantial than that in the discrete tie odels; see for exaple, PY 5 and Yu. Considering the following Vasicek odel with y = : dy t = κydt + σdw t. 3. The paraeter of interest here is κ that captures the persistence of the process. When κ >, the process is stationary and κ deterines the speed of ean reversion. The observed data are assued to be recorded discretely at h, h,, nh= T in the tie interval, T ]. So h is the saple interval, n is the total nuber of observations and T is the tie span. In this case, Yu showed that the bias of the ML estiator of κ is: E κ κ = 3 + e κh + ot. 3. T It is clear fro 3. that the bias depends on T, κ, and h. When κ is close to, or h is close to, which is epirically realistic, the bias is about /T. When T is not too big, this bias is very big, relative to the true value of κ. The result can be extended to the Vasicek odel with an unknown ean and to the square root odel; see Tang and Chen 9. The large bias otivated PY 5 to use the jackknife ethod. When κ =, the odel has a unit root, and the exact discrete tie representation is a rando walk. In this case, it can be shown that the bias of the ML estiator of κ is:.78 T + ot. 3.3a The bias forula for κ is siilar to that for β in.. However, the direction of the bias is opposite and the bias depends on T, not n. When T is not big, this bias is very big relative to the true value

15 of κ. If T, the so-called long span liit distribution of κ is given by: T κ W dw W. See, for exaple, Phillips 987b and Zhou and Yu. Siilarly, the long span liit distributions of the sub-saple estiators are: T κ j j/ j / W dw T, j =,...,. j/ W j / Obviously, the only difference between these liit distributions and those in the discrete tie unit root odel is the inus sign. Hence, the variances and covariances of the two sets of liit distribution are the sae but the expected values of the change the sign. Of course, the rate of convergence changes fro n to T. Consequently, we have the following theore. Theore 3. When =, the approxiate variance for the full saple estiator and the first subsaple estiator in the continuous tie unit root odel is: T V ar κ = T V ar κ =.3 + OT. 3. Siilarly, the approxiate variance for the second subsaple estiator in the continuous tie unit root odel is: T V ar κ = OT. 3.5 The approxiate covariances between the full saple estiator and the two subsaple estiators are: T Cov κ, κ = OT ; 3.6 T Cov κ, κ = OT ; 3.7 T Cov κ, κ =. + OT. 3.8 The optial jackknife estiator of κ in the continuous tie unit root odel when = is: κ CY =.839 κ.677 κ +.69 κ. 3.9 Reark 3. The optial jackknife estiator of κ in the continuous tie unit root odel has the sae weights and the expression as that of β in the discrete tie unit root odel. This is not surprising because there are only two differences in the liit theory, the sign of the bias and the rate of convergence. These differences do not have any ipact on the weights as they are canceled out. 5

16 Reark 3.3 The effect of the initial condition on the bias in the continuous tie odel was recently pointed out in Yu. When the initial value is in Model 3. with κ >, Yu showed that the bias of the ML estiator of κ is: E κ κ = 3 + e κh + ot. 3. T However, when the initial value is N, σ /κ in Model 3. with κ, the bias of the ML estiator of κ is: Monte Carlo Studies E κ κ = 3 + e κh e nκh T T n e κh + ot. 3. In this section, we check the finite saple properties of the proposed jackknife against the CK jackknife using siulated data. This is iportant for two reasons. First, it is iportant to easure the effi - ciency gain of the proposed ethod. Second, the weights are obtained based on the variances and the covariances of the liit distributions but not based on the finite saple distributions. These approxiations ay or ay not have ipacts on the optial weights. Hence, it is inforative to exaine the iportance of the approxiation error. Although it is diffi cult to obtain the analytical expressions for the variances and the covariances of the finite saple distribution, we can copute the fro siulated data in a Monte Carlo study, provided the nuber of replications is large enough. In this paper, we always set the nuber of replications at 5,.. Discrete Tie Unit Root Model First, we siulate data fro the following discrete tie unit root odel with initial value y = : y t = y t + ε t, ε t iid N,, t =,,..., n. We evaluate the perforance of alternative jackknife ethods by applying the to five different saple sizes, i.e. n =,, 8, 96, 8. In particular, we copare three ethods when =, the CK jackknife ethod based on.9, the CY jackknife ethod based on.8 where the weights are derived fro the approxiate variances and covariances, the CY jackknife ethod where the weights are calculated fro the exact variances and covariances obtained fro the finite saple distributions. The results are reported in Table, where for each case we calculate the ean, the variance and the RMSE for the estiates of β, all across 5, saple paths. In addition, we calculate the ratio of the variances and the RMSEs for the CK estiates and the proposed CY estiates to easure the effi ciency loss in using the CK estiate. 6

17 Table : Finite saple perforance of alternative jackknife estiators for the discrete tie unit root odel when = n Statistics CK CY Effi ciency of Exact CY CK relative to CY Bias Var* RMSE* Bias Var* RMSE* Bias Var* RMSE* Bias Var* RMSE* Bias Var* RMSE* Several interesting results eerge fro the table. First, the bias in the estiate of β for the CK ethod and the CY ethod is very siilar in every case. Second, the variance in the estiate of β for the CY ethod is significantly saller than that for the CK ethod in each case. The reduction in the variance is over % in all cases. Consequently, the RMSE is saller for the CY ethod. Third, although the exact CY ethod provides a saller variance, the difference between the two CY ethods is so sall, suggesting that the proposed CY works well, and it is not necessary to bear the additional coputational cost associated with calculating the variances and covariances fro the finite saple distributions. It is worth pointing out that even when n =, a very sall saple size that could have bigger iplications for the approxiation error, the difference between the two CY ethods is still negligible in ters of the bias, variance and RMSE. To understand why the two CY estiates are so siilar, Table 5 reports the weights obtained fro the finite saple distributions. For the purpose of coparison, we also report the weights obtained fro the liit distribution. It can be clearly seen that when n, all the weights converge. When n is as sall as, the weights are not very different fro those obtained when n is infinite.. Continuous Tie Unit Root Model Second, we siulate data fro the following continuous tie unit root odel with initial value y = : dy t = κydt + dw t, 7

18 Table 5: Jackknife weights based on the finite saple distributions and the liit distribution for the discrete tie unit root odel. n b CY a CY a CY Table 6: Finite saple perforance of alternative jackknife estiators for the continuous tie unit root odel when =. n h T Statistics CK CY Effi ciency of CK Exact CY relative to CY /5 Bias Var RMSE /5 5 Bias Var RMSE /5 Bias Var RMSE /5 5 Bias Var RMSE with κ =. Here, we generate data based on its exact discrete for as y t+h = y t +ε t, with ε t N, h. As for the discrete tie odel, we eploy the three jackknife ethod to estiate κ. Table 6 shows the results based on two sapling intervals h = /5, /5, corresponding to the weekly and daily frequency. The tie span, T, is set at years and 5 years. These settings are epirically realistic for odeling interest rates and volatility. Several interesting results also eerge fro the table. First. the bias in the estiate of κ for the CK ethod and the CY ethod is very siilar in each case. Second, the variance in the estiate of κ for the CY ethod is always significantly saller than those for the CK ethod. The reduction in the variance is at least %. Consequently, the RMSE is saller for the CY ethod. Third, the difference between the two CY ethods is very sall, suggesting that the proposed CY works well and it is not necessary to bear the additional coputational cost associated with calculating the variances and covariances fro the finite saple distributions. 8

19 Table 7: Weights based on the finite saple distributions and the liit distribution n h T b CY a CY a CY / / / / fixed To understand why the two CY estiates are so siilar in the continuous tie odel, Table 7 reports the weights obtained fro the finite saple distributions. For the purpose of coparison, we also report the weights obtained fro the liit distribution. It can be clearly seen, as T but not n increases, all the weights get close to those obtained fro the liit distribution. When T is as sall as, the weights are not different fro those obtained when T is infinite. 5 Conclusion This paper has introduced a new jackknife procedure for unit root odels that offers iproveent over the jackknife ethodology of CK. The proposed estiator is optial in the sense that it iniizes the variance of the jackknife estiator while aintaining the desirable property of being able to reove the first order bias. The new ethod is applicable to both discrete tie and continuous tie unit root odels. Siulation studies have shown that this new ethod reduces the variance by ore than % relative to the estiator of CK without coproising the bias. Although it is not pursued in the present paper, it ay be useful to further iprove our ethod by using alternative values for, and hence, further reduce the variance and RMSE. Furtherore, there are other odels where the asyptotic theory is critically dependent on the initial condition. Exaples would include near unit root odels and explosive processes. It ay be interesting to extend the results in the present paper to these odels. We plan to report these results in our later work. A Proof of Proposition. APPENDIX The notations in the proposition are defined as follows. H = Hθ = λb a] λ θ + aθ λ ] λb a]; 9

20 H = H, i.e.: H = λb a] + aλ λb a]; H denotes the first derivative of Hθ with respect to θ, evaluated at θ =, i.e.: H = H θ== λb a]. θ λ Siilarly, H denotes the second derivative of Hθ with respect to θ, evaluated at θ =, i.e.: H = H θ θ== a λ λb a]. The first derivative of M a,b with respect to θ can be written as: M a,b = b a exp θ ] b a H / θ exp θ ] b a 3/ H H. θ The second derivative of M a,b with respect to θ is: M a,b θ = b a exp θ b a + 3 exp θ b a ] H / + exp θ b a ] H 5/ H θ ] H 3/ b a H ] H θ θ Setting θ = and by standard nuerical integration techniques, we can obtain the results in Proposition.. B Proof of Lea.5 Taking the derivative of MGF with respect to θ, we get: M a,b,c,d θ, θ, ϕ, ϕ θ = E Na, b expθ Na, b θ Da, b + ϕ Nc, d ϕ Dc, d]. Setting θ =, taking the derivative with respect to ϕ, and then evaluating it at ϕ =, we have, { ] } Ma,b,c,d θ, θ, ϕ, ϕ θ= / ϕ θ ϕ == E {Na, bnc, d exp θ Da, b ϕ Dc, d]}.

21 Consequently, = = E = E { ] } Ma,b,c,d θ, θ, ϕ, ϕ θ= / ϕ θ ϕ = dθ dϕ E {Na, bnc, d exp θ Da, b ϕ Dc, d]} dθ dϕ Nc, d exp ϕ Dc, d]dϕ Na, b exp θ Da, b]dθ ] Na, b Nc, d. Da, b Dc, d C Proof of Proposition.6 To prove this proposition, we follow CK. STEP : Deriving the MGF for E W dw / W dw W / W Let Xt and Y t t, ] be the OU process defined by. dxt = γxtdt + dw t, X =, dy t = λy tdt + dw t, Y =. The easures induced by X and Y are denoted as µ X and µ Y, respectively. By Girsanov s Theore, dµ X s = exp γ λ dµ Y stdst γ λ ] st dt, where the left side is the Radon-Nikody derivative evaluated at st. Let the MGF of E W dw / W dw be M W,,,/ θ, θ, ϕ, ϕ so that M,,,/ θ, θ, ϕ, ϕ = E / W exp / ] / θ W dw + θ W + ϕ W dw + ϕ W. With the change of easure, we have the following forula by setting γ = : M,,,/ θ, θ, ϕ, ϕ / / = E exp θ Y dy + θ Y + ϕ Y dy + ϕ Y λ Y dy + λ Y ].

22 By Ito s calculus, we get: M,,,/ θ, θ, ϕ, ϕ = exp θ ϕ + λ { θ λ E exp = exp θ ϕ + λ Y + ϕ { E exp / Y + ϕ + θ + λ Y + θ λ Y + ϕ ]} / Y + ϕ Y ]} θ + λ Y / where λ = θ. To calculate the expectation, we first take the expectation of M,,,/ θ, θ, ϕ, ϕ conditional on Ϝ /, which is the siga field generated by W on, ], and then change the easure by Girsanov s Theore with introducing another OU process. Note that, M,,,/ θ, θ, ϕ, ϕ, Ϝ / = E = exp θ ϕ + λ ϕ exp / Y + ϕ M,,,/ θ, θ, ϕ, ϕ Ϝ / ] { Y E exp ] θ λ Y Y ]}. The MGF of the noncentral χ distribution, θ λ Y, is θ λϖ ] / exp θ λ ζy ] with ϖ = expλ λ and ζ = θ λϖ ] expλ. Therefore, M,,,/ θ, θ, ϕ, ϕ, Ϝ / = θ λϖ ] / exp θ ϕ + λ { θ λ E exp ζ + ϕ ] Y } / + ϕ Y. Now we introduce another process Zt on, ], given by dzt = ηztdt + dw t, Z =. By changing the easure and setting ϕ = λ η / to cancel out / Z, we have { θ λ E exp ζ + ϕ ] } / Y + ϕ Y { θ λ = E exp ζ + ϕ ] } / Z + λ η ZdZ ] η λ = exp E exp τz,

23 with τ = θ λ ζ + ϕ + λ η. Considering Z N, ϖ z where ϖ z = expη η, we have E exp { θ λ ζ + ϕ ] Y Finally by cobining the equations above, we obtain M,,,/ θ, θ, ϕ, ϕ = Note that, λ θ λϖ ] exp and that η τϖ z] exp = exp λ = exp η where π = θ λζ + ϕ + λ. Thus, } / + ϕ Y = τϖ z] / exp η λ { } / η + λ θ λϖ ] τϖ z] exp exp θ ϕ. + exp λ θ λ + exp η π η exp λ exp η exp λ =. λ θ λ exp η η = π η η, M,,,/ θ, θ, ϕ, ϕ = H,,,/ θ, θ, ϕ, ϕ / exp θ ϕ, λ, with H,,,/ θ, θ, ϕ, ϕ = λ θ λ ] λ η π η ] η. ] Ma,b,c,d θ STEP : Copute, θ,ϕ, ϕ θ θ= / ϕ ϕ =. Denote H,,,/ θ, θ, ϕ, ϕ and M,,,/ θ, θ, ϕ, ϕ by H and M. Note that M = H / exp θ ϕ where λ = θ and η = θ + ϕ. Taking the partial derivative with respect to θ and evaluating it at θ =, we have, M θ== θ exp ϕ H / exp ϕ H 3/ H θ=. θ Then taking the second derivative with respect to ϕ and evaluating it at ϕ =, we obtain ] M θ θ= ϕ = ϕ = 8 H / + H 3/ H ϕ = + H θ=,ϕ ϕ 8 θ = + 3 H H H 5/ ϕ = θ=,ϕ ϕ θ =, H θ θ= ϕ = ϕ 3

24 fro which we require the following four expressions H = H θ=,ϕ == H ϕ H θ θ=,ϕ == λ λ η + λ η λ η ; ϕ == η λ λ η ; η η η H θ θ= ϕ == λ η ϕ λη. λ ; Finally, we have the following expression that is aenable for nuerical integrations. ] M θ θ= ϕ = ϕ = 8 H / + H 3/ + 3 H 5/ { η η λ 8λ λ η ] λ λ 3 η 8η λ + η η η λ λη ] λ. η } By the sae arguent, we have the MGFs of M,,/, θ, θ, ϕ, ϕ and M,/,/, θ, θ, ϕ, ϕ. To distinguish λ, η and π in the three cases, we use subscripts. To su up: The MGF M,,,/ θ, θ, ϕ, ϕ is given by { λ,,,/ exp θ ϕ, θ λ,,,/ λ,,,/ ] η,,,/ π,,,/ η ]} / η,,,/ where λ,,,/ = θ, η,,,/ = θ ϕ, and π,,,/ = θ λ,,,/ ζ,,,/ + ϕ + λ,,,/ with ζ,,,/ = exp λ,,,/ θ λ,,,/ ϖ ], ϖ = exp,,,/,,,/ τ,,,/ = θ λ,,,/ ζ,,,/ + ϕ + λ,,,/ η,,,/, and ϖ z,,,,/ = exp The MGF M,,/, θ, θ, ϕ, ϕ is given by { λ,,/, exp θ ϕ, θ + ϕ λ,,/, λ,,/, ] η,,/, η,,,/ η.,,,/ π,,/, η,,/, λ,,,/ λ,,,,/ η,,/, ]} /

25 where λ,,/, = θ ϕ, η = θ and π,,/,,,/, = θ + ϕ λ,,/, ζ ] exp λ,,/, λ,,/, with ζ =,,/, θ + ϕ λ,,/, ϖ,,/, τ,,/, = θ+ϕ λ,,/, ζ ϕ,,/, + λ η,,/,,,/,, and ϖ z, = exp,,/, The MGF M,/,/, θ, θ, ϕ, ϕ is given by { λ,/,/, exp θ ϕ where ϖ, λ = exp,/,/, ϕ λ,/,/, λ,/,/, λ,/,/, ζ,/,/, +θ ϕ +λ,/,/, with ζ,/,/, = ] η,/,/,, ϖ,,/,,,/, = exp η,,/, η.,,/, π,/,/, η,/,/, ϕ + λ,,/, λ,,,/, η,/,/, λ, λ,/,/, = ϕ, η,/,/, = θ, and π,/,/, = ϕ,/,/, θ λ,/,/, ϖ,/,/,], expλ,/,/, τ,/,/, = ϕ λ,/,/, ζ,/,/, + θ ϕ + λ,/,/, η,/,/,, and ϖ z = expη,/,/, η,/,/,. The coputation of the second derivative of the MGF siply follows step. For exaple, the second derivative of MGF M,,,/ θ, θ, ϕ, ϕ is M,,,/θ, θ,ϕ, ϕ θ θ= ϕ = ϕ = 8 H /,,,,/ + H 3/,,,,/ { 3 8η,,,/ + 3 H 5/,,,,/ λ,,,/ η,,,/ η,,,/ λ,,,/ 8λ,,,/ λ,,,/ η,,,/ λ,,,/ and H,,,,/ = θ and η = θ + ϕ,,,/. η,,,/ η,,,/ λ,,,/ η,,,/ λ,,,/ η,,,/ ] λ,,,/ η,,,/ + λ,,,/ η,,,/ η,,,/ η,,,/ λ,,,/ λ,,,/ λ,,,/ η,,,/ ]. } with λ,,,/ = ]} / 5

26 The second derivative of MGF M,,/, θ, θ, ϕ, ϕ is M,,/,θ, θ,ϕ, ϕ θ θ= ϕ = ϕ = 8 H /,,,/, + H 3/,,,/, { 3 8η,,/, + 3 H 5/,,,/, λ,,/, η,,/, λ,,/, λ,,/, λ,,,/ 8λ,,/, λ,,/, λ,,/, and H,,,,/ = θ + ϕ and η =,,,/ θ. λ,,/, + η,,/, η,,,/ η,,/, λ,,/, η,,/, ] η,,/, η,,/, + λ,,,/ η,,,/ λ,,/, η,,/, η ] λ,,/,,,/,, η,,,/ The second derivative of MGF M,/,/, θ, θ, ϕ, ϕ is ] M,/,/, θ, θ, ϕ, ϕ ϕ = ϕ = 6 H /,,/,/, + H 3/,,/,/, { 8η,/,/, + 3 H 5/,,/,/, η,/,/, and H,,/,/, = η,/,/, λ,/,/, η,/,/, λ,/,/, 8λ,/,/, λ,/,/, λ,/,/, with λ,/,/, = ϕ and η,/,/, = θ. λ,/,/, λ,/,/, η,/,/, ], λ,/,/, η,/,/, ] η,/,/, + λ,/,/, η,/,/, η,/,/, λ,,,/ η,/,/, λ,/,/, } with λ,,,/ = λ,/,/, η,/,/, } D Proof of Theore.9 Using the results fro Propositions. and.6, and substituting the values in Equation.8, we have the stated result. References 6

27 Chabers, M. J., Jackknife estiation and inference in stationary autoregressive odels, Departent of Econoics, University of Essex, Working paper. Chabers, M.J., and M. Kyriacou,, Jackknife bias reduction in the presence of a unit root, Departent of Econoics, University of Essex, Working paper. Chiquoine, B., and E. Hjalarsson, 9, Jackknifing stock return predictions, Journal of Epirical Finance, 6, , Miller, R.G, 97, The jackknife - A review, Bioetrika, 6, -5. Hahn, J. and W. K. Newey,, Jackknife and analytical bias reduction for nonlinear panel odels. Econoetrica, 7, Phillips, P.C.B., 987a, Tie Series Regression with a Unit Root, Econoetrica, 55, Phillips, P.C.B., 987b, Toward a unified asyptotic theory for autoregression, Bioetrika, 7, Phillips, P.C.B., and T. Magdalinos, 9, Unit root and cointegrating liit theory when initialization is in the infinite past, Econoetric Theory, 5, Phillips, P.C.B., and J. Yu, 5, Jackknifing bond option prices, Review of Financial Studies, 8, Quenouille, M. H., 99, Approxiate Tests of Correlation in Tie-Series, Journal of the Royal Statistical Society. Series B,, Quenouille, M. H., 956, Notes on bias in estiation, Bioetrika 3, Sargan, J. D., 976, Econoetric Estiators and the Edgeworth Approxiation, Econoetrica,, 8. Schucany, W. R., Gray, H. L., Owen, D. B., 97, On bias reduction in estiation, Journal of the Aerican Statistical Association, 66, Shenton L.R., and W. L. Johnson, 965, Moents of a serial correlation coeffi cient, Journal of the Royal Statistical Society, 7, Tang, C. Y., and S. X. Chen., 9. Paraeter estiation and bias correction for diffusion processes. Journal of Econoetrics. 9, 65-8 Tukey, J. W., 958, Bias and confidence in not-quite large saples, Annals of Matheatical Statistics, 9, 6. White, J. S., 96, Asyptotic Expansions for the Mean and Variance of the Serial Correlation Coeffi cient, Bioetrika, 8, Wu, C. F. J., 986, Jackknife, bootstrap, and other resapling ethods in regression analysis, The Annals of Statistics,, Yu, J.,, Bias in the Estiation of the Mean Reversion Paraeter in Continuous Tie Models, Journal of Econoetrics, forthcoing. Zhou, Q. and J. Yu,, Asyptotic distributions of the least squares estiator for diffusion process, Singapore Manageent University, Working Paper;. 7

Optimal Jackknife for Unit Root Models

Optimal Jackknife for Unit Root Models Optimal Jackknife for Unit Root Models Ye Chen and Jun Yu Singapore Management University October 19, 2014 Abstract A new jackknife method is introduced to remove the first order bias in the discrete time

More information

Keywords: Estimator, Bias, Mean-squared error, normality, generalized Pareto distribution

Keywords: Estimator, Bias, Mean-squared error, normality, generalized Pareto distribution Testing approxiate norality of an estiator using the estiated MSE and bias with an application to the shape paraeter of the generalized Pareto distribution J. Martin van Zyl Abstract In this work the norality

More information

Bootstrapping Dependent Data

Bootstrapping Dependent Data Bootstrapping Dependent Data One of the key issues confronting bootstrap resapling approxiations is how to deal with dependent data. Consider a sequence fx t g n t= of dependent rando variables. Clearly

More information

Biostatistics Department Technical Report

Biostatistics Department Technical Report Biostatistics Departent Technical Report BST006-00 Estiation of Prevalence by Pool Screening With Equal Sized Pools and a egative Binoial Sapling Model Charles R. Katholi, Ph.D. Eeritus Professor Departent

More information

Testing equality of variances for multiple univariate normal populations

Testing equality of variances for multiple univariate normal populations University of Wollongong Research Online Centre for Statistical & Survey Methodology Working Paper Series Faculty of Engineering and Inforation Sciences 0 esting equality of variances for ultiple univariate

More information

A Simple Regression Problem

A Simple Regression Problem A Siple Regression Proble R. M. Castro March 23, 2 In this brief note a siple regression proble will be introduced, illustrating clearly the bias-variance tradeoff. Let Y i f(x i ) + W i, i,..., n, where

More information

A Jackknife Correction to a Test for Cointegration Rank

A Jackknife Correction to a Test for Cointegration Rank Econoetrics 205, 3, 355-375; doi:0.3390/econoetrics3020355 OPEN ACCESS econoetrics ISSN 2225-46 www.dpi.co/journal/econoetrics Article A Jackknife Correction to a Test for Cointegration Rank Marcus J.

More information

Machine Learning Basics: Estimators, Bias and Variance

Machine Learning Basics: Estimators, Bias and Variance Machine Learning Basics: Estiators, Bias and Variance Sargur N. srihari@cedar.buffalo.edu This is part of lecture slides on Deep Learning: http://www.cedar.buffalo.edu/~srihari/cse676 1 Topics in Basics

More information

DEPARTMENT OF ECONOMETRICS AND BUSINESS STATISTICS

DEPARTMENT OF ECONOMETRICS AND BUSINESS STATISTICS ISSN 1440-771X AUSTRALIA DEPARTMENT OF ECONOMETRICS AND BUSINESS STATISTICS An Iproved Method for Bandwidth Selection When Estiating ROC Curves Peter G Hall and Rob J Hyndan Working Paper 11/00 An iproved

More information

TEST OF HOMOGENEITY OF PARALLEL SAMPLES FROM LOGNORMAL POPULATIONS WITH UNEQUAL VARIANCES

TEST OF HOMOGENEITY OF PARALLEL SAMPLES FROM LOGNORMAL POPULATIONS WITH UNEQUAL VARIANCES TEST OF HOMOGENEITY OF PARALLEL SAMPLES FROM LOGNORMAL POPULATIONS WITH UNEQUAL VARIANCES S. E. Ahed, R. J. Tokins and A. I. Volodin Departent of Matheatics and Statistics University of Regina Regina,

More information

ESTIMATING AND FORMING CONFIDENCE INTERVALS FOR EXTREMA OF RANDOM POLYNOMIALS. A Thesis. Presented to. The Faculty of the Department of Mathematics

ESTIMATING AND FORMING CONFIDENCE INTERVALS FOR EXTREMA OF RANDOM POLYNOMIALS. A Thesis. Presented to. The Faculty of the Department of Mathematics ESTIMATING AND FORMING CONFIDENCE INTERVALS FOR EXTREMA OF RANDOM POLYNOMIALS A Thesis Presented to The Faculty of the Departent of Matheatics San Jose State University In Partial Fulfillent of the Requireents

More information

Some Proofs: This section provides proofs of some theoretical results in section 3.

Some Proofs: This section provides proofs of some theoretical results in section 3. Testing Jups via False Discovery Rate Control Yu-Min Yen. Institute of Econoics, Acadeia Sinica, Taipei, Taiwan. E-ail: YMYEN@econ.sinica.edu.tw. SUPPLEMENTARY MATERIALS Suppleentary Materials contain

More information

Testing the lag length of vector autoregressive models: A power comparison between portmanteau and Lagrange multiplier tests

Testing the lag length of vector autoregressive models: A power comparison between portmanteau and Lagrange multiplier tests Working Papers 2017-03 Testing the lag length of vector autoregressive odels: A power coparison between portanteau and Lagrange ultiplier tests Raja Ben Hajria National Engineering School, University of

More information

Probability Distributions

Probability Distributions Probability Distributions In Chapter, we ephasized the central role played by probability theory in the solution of pattern recognition probles. We turn now to an exploration of soe particular exaples

More information

Estimation of the Mean of the Exponential Distribution Using Maximum Ranked Set Sampling with Unequal Samples

Estimation of the Mean of the Exponential Distribution Using Maximum Ranked Set Sampling with Unequal Samples Open Journal of Statistics, 4, 4, 64-649 Published Online Septeber 4 in SciRes http//wwwscirporg/ournal/os http//ddoiorg/436/os4486 Estiation of the Mean of the Eponential Distribution Using Maiu Ranked

More information

Non-Parametric Non-Line-of-Sight Identification 1

Non-Parametric Non-Line-of-Sight Identification 1 Non-Paraetric Non-Line-of-Sight Identification Sinan Gezici, Hisashi Kobayashi and H. Vincent Poor Departent of Electrical Engineering School of Engineering and Applied Science Princeton University, Princeton,

More information

Block designs and statistics

Block designs and statistics Bloc designs and statistics Notes for Math 447 May 3, 2011 The ain paraeters of a bloc design are nuber of varieties v, bloc size, nuber of blocs b. A design is built on a set of v eleents. Each eleent

More information

Extension of CSRSM for the Parametric Study of the Face Stability of Pressurized Tunnels

Extension of CSRSM for the Parametric Study of the Face Stability of Pressurized Tunnels Extension of CSRSM for the Paraetric Study of the Face Stability of Pressurized Tunnels Guilhe Mollon 1, Daniel Dias 2, and Abdul-Haid Soubra 3, M.ASCE 1 LGCIE, INSA Lyon, Université de Lyon, Doaine scientifique

More information

arxiv: v1 [cs.ds] 3 Feb 2014

arxiv: v1 [cs.ds] 3 Feb 2014 arxiv:40.043v [cs.ds] 3 Feb 04 A Bound on the Expected Optiality of Rando Feasible Solutions to Cobinatorial Optiization Probles Evan A. Sultani The Johns Hopins University APL evan@sultani.co http://www.sultani.co/

More information

Covariance-based orthogonality tests for regressors with unknown persistence

Covariance-based orthogonality tests for regressors with unknown persistence Covariance-based orthogonality tests for regressors with unknown persistence Alex Maynard and Katsui Shiotsu Departent of Econoics, University of Toronto and Departent of Econoics, Queen s University April

More information

IN modern society that various systems have become more

IN modern society that various systems have become more Developent of Reliability Function in -Coponent Standby Redundant Syste with Priority Based on Maxiu Entropy Principle Ryosuke Hirata, Ikuo Arizono, Ryosuke Toohiro, Satoshi Oigawa, and Yasuhiko Takeoto

More information

Sharp Time Data Tradeoffs for Linear Inverse Problems

Sharp Time Data Tradeoffs for Linear Inverse Problems Sharp Tie Data Tradeoffs for Linear Inverse Probles Saet Oyak Benjain Recht Mahdi Soltanolkotabi January 016 Abstract In this paper we characterize sharp tie-data tradeoffs for optiization probles used

More information

In this chapter, we consider several graph-theoretic and probabilistic models

In this chapter, we consider several graph-theoretic and probabilistic models THREE ONE GRAPH-THEORETIC AND STATISTICAL MODELS 3.1 INTRODUCTION In this chapter, we consider several graph-theoretic and probabilistic odels for a social network, which we do under different assuptions

More information

13.2 Fully Polynomial Randomized Approximation Scheme for Permanent of Random 0-1 Matrices

13.2 Fully Polynomial Randomized Approximation Scheme for Permanent of Random 0-1 Matrices CS71 Randoness & Coputation Spring 018 Instructor: Alistair Sinclair Lecture 13: February 7 Disclaier: These notes have not been subjected to the usual scrutiny accorded to foral publications. They ay

More information

A Note on the Applied Use of MDL Approximations

A Note on the Applied Use of MDL Approximations A Note on the Applied Use of MDL Approxiations Daniel J. Navarro Departent of Psychology Ohio State University Abstract An applied proble is discussed in which two nested psychological odels of retention

More information

An Extension to the Tactical Planning Model for a Job Shop: Continuous-Time Control

An Extension to the Tactical Planning Model for a Job Shop: Continuous-Time Control An Extension to the Tactical Planning Model for a Job Shop: Continuous-Tie Control Chee Chong. Teo, Rohit Bhatnagar, and Stephen C. Graves Singapore-MIT Alliance, Nanyang Technological Univ., and Massachusetts

More information

AN OPTIMAL SHRINKAGE FACTOR IN PREDICTION OF ORDERED RANDOM EFFECTS

AN OPTIMAL SHRINKAGE FACTOR IN PREDICTION OF ORDERED RANDOM EFFECTS Statistica Sinica 6 016, 1709-178 doi:http://dx.doi.org/10.5705/ss.0014.0034 AN OPTIMAL SHRINKAGE FACTOR IN PREDICTION OF ORDERED RANDOM EFFECTS Nilabja Guha 1, Anindya Roy, Yaakov Malinovsky and Gauri

More information

The proofs of Theorem 1-3 are along the lines of Wied and Galeano (2013).

The proofs of Theorem 1-3 are along the lines of Wied and Galeano (2013). A Appendix: Proofs The proofs of Theore 1-3 are along the lines of Wied and Galeano (2013) Proof of Theore 1 Let D[d 1, d 2 ] be the space of càdlàg functions on the interval [d 1, d 2 ] equipped with

More information

Pseudo-marginal Metropolis-Hastings: a simple explanation and (partial) review of theory

Pseudo-marginal Metropolis-Hastings: a simple explanation and (partial) review of theory Pseudo-arginal Metropolis-Hastings: a siple explanation and (partial) review of theory Chris Sherlock Motivation Iagine a stochastic process V which arises fro soe distribution with density p(v θ ). Iagine

More information

2 Q 10. Likewise, in case of multiple particles, the corresponding density in 2 must be averaged over all

2 Q 10. Likewise, in case of multiple particles, the corresponding density in 2 must be averaged over all Lecture 6 Introduction to kinetic theory of plasa waves Introduction to kinetic theory So far we have been odeling plasa dynaics using fluid equations. The assuption has been that the pressure can be either

More information

Physically Based Modeling CS Notes Spring 1997 Particle Collision and Contact

Physically Based Modeling CS Notes Spring 1997 Particle Collision and Contact Physically Based Modeling CS 15-863 Notes Spring 1997 Particle Collision and Contact 1 Collisions with Springs Suppose we wanted to ipleent a particle siulator with a floor : a solid horizontal plane which

More information

An Approximate Model for the Theoretical Prediction of the Velocity Increase in the Intermediate Ballistics Period

An Approximate Model for the Theoretical Prediction of the Velocity Increase in the Intermediate Ballistics Period An Approxiate Model for the Theoretical Prediction of the Velocity... 77 Central European Journal of Energetic Materials, 205, 2(), 77-88 ISSN 2353-843 An Approxiate Model for the Theoretical Prediction

More information

Nonlinear Log-Periodogram Regression for Perturbed Fractional Processes

Nonlinear Log-Periodogram Regression for Perturbed Fractional Processes Nonlinear Log-Periodogra Regression for Perturbed Fractional Processes Yixiao Sun Departent of Econoics Yale University Peter C. B. Phillips Cowles Foundation for Research in Econoics Yale University First

More information

Experimental Design For Model Discrimination And Precise Parameter Estimation In WDS Analysis

Experimental Design For Model Discrimination And Precise Parameter Estimation In WDS Analysis City University of New York (CUNY) CUNY Acadeic Works International Conference on Hydroinforatics 8-1-2014 Experiental Design For Model Discriination And Precise Paraeter Estiation In WDS Analysis Giovanna

More information

AN EFFICIENT CLASS OF CHAIN ESTIMATORS OF POPULATION VARIANCE UNDER SUB-SAMPLING SCHEME

AN EFFICIENT CLASS OF CHAIN ESTIMATORS OF POPULATION VARIANCE UNDER SUB-SAMPLING SCHEME J. Japan Statist. Soc. Vol. 35 No. 005 73 86 AN EFFICIENT CLASS OF CHAIN ESTIMATORS OF POPULATION VARIANCE UNDER SUB-SAMPLING SCHEME H. S. Jhajj*, M. K. Shara* and Lovleen Kuar Grover** For estiating the

More information

Inspection; structural health monitoring; reliability; Bayesian analysis; updating; decision analysis; value of information

Inspection; structural health monitoring; reliability; Bayesian analysis; updating; decision analysis; value of information Cite as: Straub D. (2014). Value of inforation analysis with structural reliability ethods. Structural Safety, 49: 75-86. Value of Inforation Analysis with Structural Reliability Methods Daniel Straub

More information

CS Lecture 13. More Maximum Likelihood

CS Lecture 13. More Maximum Likelihood CS 6347 Lecture 13 More Maxiu Likelihood Recap Last tie: Introduction to axiu likelihood estiation MLE for Bayesian networks Optial CPTs correspond to epirical counts Today: MLE for CRFs 2 Maxiu Likelihood

More information

Modeling the Structural Shifts in Real Exchange Rate with Cubic Spline Regression (CSR). Turkey

Modeling the Structural Shifts in Real Exchange Rate with Cubic Spline Regression (CSR). Turkey International Journal of Business and Social Science Vol. 2 No. 17 www.ijbssnet.co Modeling the Structural Shifts in Real Exchange Rate with Cubic Spline Regression (CSR). Turkey 1987-2008 Dr. Bahar BERBEROĞLU

More information

W-BASED VS LATENT VARIABLES SPATIAL AUTOREGRESSIVE MODELS: EVIDENCE FROM MONTE CARLO SIMULATIONS

W-BASED VS LATENT VARIABLES SPATIAL AUTOREGRESSIVE MODELS: EVIDENCE FROM MONTE CARLO SIMULATIONS W-BASED VS LATENT VARIABLES SPATIAL AUTOREGRESSIVE MODELS: EVIDENCE FROM MONTE CARLO SIMULATIONS. Introduction When it coes to applying econoetric odels to analyze georeferenced data, researchers are well

More information

General Properties of Radiation Detectors Supplements

General Properties of Radiation Detectors Supplements Phys. 649: Nuclear Techniques Physics Departent Yarouk University Chapter 4: General Properties of Radiation Detectors Suppleents Dr. Nidal M. Ershaidat Overview Phys. 649: Nuclear Techniques Physics Departent

More information

e-companion ONLY AVAILABLE IN ELECTRONIC FORM

e-companion ONLY AVAILABLE IN ELECTRONIC FORM OPERATIONS RESEARCH doi 10.1287/opre.1070.0427ec pp. ec1 ec5 e-copanion ONLY AVAILABLE IN ELECTRONIC FORM infors 07 INFORMS Electronic Copanion A Learning Approach for Interactive Marketing to a Custoer

More information

Polygonal Designs: Existence and Construction

Polygonal Designs: Existence and Construction Polygonal Designs: Existence and Construction John Hegean Departent of Matheatics, Stanford University, Stanford, CA 9405 Jeff Langford Departent of Matheatics, Drake University, Des Moines, IA 5011 G

More information

Soft Computing Techniques Help Assign Weights to Different Factors in Vulnerability Analysis

Soft Computing Techniques Help Assign Weights to Different Factors in Vulnerability Analysis Soft Coputing Techniques Help Assign Weights to Different Factors in Vulnerability Analysis Beverly Rivera 1,2, Irbis Gallegos 1, and Vladik Kreinovich 2 1 Regional Cyber and Energy Security Center RCES

More information

MSEC MODELING OF DEGRADATION PROCESSES TO OBTAIN AN OPTIMAL SOLUTION FOR MAINTENANCE AND PERFORMANCE

MSEC MODELING OF DEGRADATION PROCESSES TO OBTAIN AN OPTIMAL SOLUTION FOR MAINTENANCE AND PERFORMANCE Proceeding of the ASME 9 International Manufacturing Science and Engineering Conference MSEC9 October 4-7, 9, West Lafayette, Indiana, USA MSEC9-8466 MODELING OF DEGRADATION PROCESSES TO OBTAIN AN OPTIMAL

More information

The Distribution of the Covariance Matrix for a Subset of Elliptical Distributions with Extension to Two Kurtosis Parameters

The Distribution of the Covariance Matrix for a Subset of Elliptical Distributions with Extension to Two Kurtosis Parameters journal of ultivariate analysis 58, 96106 (1996) article no. 0041 The Distribution of the Covariance Matrix for a Subset of Elliptical Distributions with Extension to Two Kurtosis Paraeters H. S. Steyn

More information

Examining the Effects of Site Selection Criteria for Evaluating the Effectiveness of Traffic Safety Countermeasures

Examining the Effects of Site Selection Criteria for Evaluating the Effectiveness of Traffic Safety Countermeasures Exaining the Effects of Site Selection Criteria for Evaluating the Effectiveness of Traffic Safety Countereasures Doinique Lord* Associate Professor and Zachry Developent Professor I Zachry Departent of

More information

ANALYTICAL INVESTIGATION AND PARAMETRIC STUDY OF LATERAL IMPACT BEHAVIOR OF PRESSURIZED PIPELINES AND INFLUENCE OF INTERNAL PRESSURE

ANALYTICAL INVESTIGATION AND PARAMETRIC STUDY OF LATERAL IMPACT BEHAVIOR OF PRESSURIZED PIPELINES AND INFLUENCE OF INTERNAL PRESSURE DRAFT Proceedings of the ASME 014 International Mechanical Engineering Congress & Exposition IMECE014 Noveber 14-0, 014, Montreal, Quebec, Canada IMECE014-36371 ANALYTICAL INVESTIGATION AND PARAMETRIC

More information

EE5900 Spring Lecture 4 IC interconnect modeling methods Zhuo Feng

EE5900 Spring Lecture 4 IC interconnect modeling methods Zhuo Feng EE59 Spring Parallel LSI AD Algoriths Lecture I interconnect odeling ethods Zhuo Feng. Z. Feng MTU EE59 So far we ve considered only tie doain analyses We ll soon see that it is soeties preferable to odel

More information

C na (1) a=l. c = CO + Clm + CZ TWO-STAGE SAMPLE DESIGN WITH SMALL CLUSTERS. 1. Introduction

C na (1) a=l. c = CO + Clm + CZ TWO-STAGE SAMPLE DESIGN WITH SMALL CLUSTERS. 1. Introduction TWO-STGE SMPLE DESIGN WITH SMLL CLUSTERS Robert G. Clark and David G. Steel School of Matheatics and pplied Statistics, University of Wollongong, NSW 5 ustralia. (robert.clark@abs.gov.au) Key Words: saple

More information

Information Loss in Volatility Measurement with Flat Price Trading 1

Information Loss in Volatility Measurement with Flat Price Trading 1 Inforation Loss in Volatility Measureent with Flat Price Trading Peter C. B. Phillips Yale University, University of Auckland, University of York & Singapore Manageent University Jun Yu Singapore Manageent

More information

Economics Division University of Southampton Southampton SO17 1BJ, UK. Title Overlapping Sub-sampling and invariance to initial conditions

Economics Division University of Southampton Southampton SO17 1BJ, UK. Title Overlapping Sub-sampling and invariance to initial conditions Economics Division University of Southampton Southampton SO17 1BJ, UK Discussion Papers in Economics and Econometrics Title Overlapping Sub-sampling and invariance to initial conditions By Maria Kyriacou

More information

Feature Extraction Techniques

Feature Extraction Techniques Feature Extraction Techniques Unsupervised Learning II Feature Extraction Unsupervised ethods can also be used to find features which can be useful for categorization. There are unsupervised ethods that

More information

Necessity of low effective dimension

Necessity of low effective dimension Necessity of low effective diension Art B. Owen Stanford University October 2002, Orig: July 2002 Abstract Practitioners have long noticed that quasi-monte Carlo ethods work very well on functions that

More information

Meta-Analytic Interval Estimation for Bivariate Correlations

Meta-Analytic Interval Estimation for Bivariate Correlations Psychological Methods 2008, Vol. 13, No. 3, 173 181 Copyright 2008 by the Aerican Psychological Association 1082-989X/08/$12.00 DOI: 10.1037/a0012868 Meta-Analytic Interval Estiation for Bivariate Correlations

More information

Tail Estimation of the Spectral Density under Fixed-Domain Asymptotics

Tail Estimation of the Spectral Density under Fixed-Domain Asymptotics Tail Estiation of the Spectral Density under Fixed-Doain Asyptotics Wei-Ying Wu, Chae Young Li and Yiin Xiao Wei-Ying Wu, Departent of Statistics & Probability Michigan State University, East Lansing,

More information

Bayesian inference for stochastic differential mixed effects models - initial steps

Bayesian inference for stochastic differential mixed effects models - initial steps Bayesian inference for stochastic differential ixed effects odels - initial steps Gavin Whitaker 2nd May 2012 Supervisors: RJB and AG Outline Mixed Effects Stochastic Differential Equations (SDEs) Bayesian

More information

The degree of a typical vertex in generalized random intersection graph models

The degree of a typical vertex in generalized random intersection graph models Discrete Matheatics 306 006 15 165 www.elsevier.co/locate/disc The degree of a typical vertex in generalized rando intersection graph odels Jerzy Jaworski a, Michał Karoński a, Dudley Stark b a Departent

More information

Generalized eigenfunctions and a Borel Theorem on the Sierpinski Gasket.

Generalized eigenfunctions and a Borel Theorem on the Sierpinski Gasket. Generalized eigenfunctions and a Borel Theore on the Sierpinski Gasket. Kasso A. Okoudjou, Luke G. Rogers, and Robert S. Strichartz May 26, 2006 1 Introduction There is a well developed theory (see [5,

More information

Lower Bounds for Quantized Matrix Completion

Lower Bounds for Quantized Matrix Completion Lower Bounds for Quantized Matrix Copletion Mary Wootters and Yaniv Plan Departent of Matheatics University of Michigan Ann Arbor, MI Eail: wootters, yplan}@uich.edu Mark A. Davenport School of Elec. &

More information

A Simplified Analytical Approach for Efficiency Evaluation of the Weaving Machines with Automatic Filling Repair

A Simplified Analytical Approach for Efficiency Evaluation of the Weaving Machines with Automatic Filling Repair Proceedings of the 6th SEAS International Conference on Siulation, Modelling and Optiization, Lisbon, Portugal, Septeber -4, 006 0 A Siplified Analytical Approach for Efficiency Evaluation of the eaving

More information

Model Fitting. CURM Background Material, Fall 2014 Dr. Doreen De Leon

Model Fitting. CURM Background Material, Fall 2014 Dr. Doreen De Leon Model Fitting CURM Background Material, Fall 014 Dr. Doreen De Leon 1 Introduction Given a set of data points, we often want to fit a selected odel or type to the data (e.g., we suspect an exponential

More information

3.3 Variational Characterization of Singular Values

3.3 Variational Characterization of Singular Values 3.3. Variational Characterization of Singular Values 61 3.3 Variational Characterization of Singular Values Since the singular values are square roots of the eigenvalues of the Heritian atrices A A and

More information

THE AVERAGE NORM OF POLYNOMIALS OF FIXED HEIGHT

THE AVERAGE NORM OF POLYNOMIALS OF FIXED HEIGHT THE AVERAGE NORM OF POLYNOMIALS OF FIXED HEIGHT PETER BORWEIN AND KWOK-KWONG STEPHEN CHOI Abstract. Let n be any integer and ( n ) X F n : a i z i : a i, ± i be the set of all polynoials of height and

More information

Proc. of the IEEE/OES Seventh Working Conference on Current Measurement Technology UNCERTAINTIES IN SEASONDE CURRENT VELOCITIES

Proc. of the IEEE/OES Seventh Working Conference on Current Measurement Technology UNCERTAINTIES IN SEASONDE CURRENT VELOCITIES Proc. of the IEEE/OES Seventh Working Conference on Current Measureent Technology UNCERTAINTIES IN SEASONDE CURRENT VELOCITIES Belinda Lipa Codar Ocean Sensors 15 La Sandra Way, Portola Valley, CA 98 blipa@pogo.co

More information

Analyzing Simulation Results

Analyzing Simulation Results Analyzing Siulation Results Dr. John Mellor-Cruey Departent of Coputer Science Rice University johnc@cs.rice.edu COMP 528 Lecture 20 31 March 2005 Topics for Today Model verification Model validation Transient

More information

Multi-Dimensional Hegselmann-Krause Dynamics

Multi-Dimensional Hegselmann-Krause Dynamics Multi-Diensional Hegselann-Krause Dynaics A. Nedić Industrial and Enterprise Systes Engineering Dept. University of Illinois Urbana, IL 680 angelia@illinois.edu B. Touri Coordinated Science Laboratory

More information

Inference in the Presence of Likelihood Monotonicity for Polytomous and Logistic Regression

Inference in the Presence of Likelihood Monotonicity for Polytomous and Logistic Regression Advances in Pure Matheatics, 206, 6, 33-34 Published Online April 206 in SciRes. http://www.scirp.org/journal/ap http://dx.doi.org/0.4236/ap.206.65024 Inference in the Presence of Likelihood Monotonicity

More information

Numerical Studies of a Nonlinear Heat Equation with Square Root Reaction Term

Numerical Studies of a Nonlinear Heat Equation with Square Root Reaction Term Nuerical Studies of a Nonlinear Heat Equation with Square Root Reaction Ter Ron Bucire, 1 Karl McMurtry, 1 Ronald E. Micens 2 1 Matheatics Departent, Occidental College, Los Angeles, California 90041 2

More information

Statistical Logic Cell Delay Analysis Using a Current-based Model

Statistical Logic Cell Delay Analysis Using a Current-based Model Statistical Logic Cell Delay Analysis Using a Current-based Model Hanif Fatei Shahin Nazarian Massoud Pedra Dept. of EE-Systes, University of Southern California, Los Angeles, CA 90089 {fatei, shahin,

More information

The Wilson Model of Cortical Neurons Richard B. Wells

The Wilson Model of Cortical Neurons Richard B. Wells The Wilson Model of Cortical Neurons Richard B. Wells I. Refineents on the odgkin-uxley Model The years since odgkin s and uxley s pioneering work have produced a nuber of derivative odgkin-uxley-like

More information

Testing for a Break in the Memory Parameter

Testing for a Break in the Memory Parameter Testing for a Break in the Meory Paraeter Fabrizio Iacone Econoics Departent London School of Econoics Houghton Street London WCA AE April 5, 005 Abstract It is often of interest in the applied analysis

More information

Symbolic Analysis as Universal Tool for Deriving Properties of Non-linear Algorithms Case study of EM Algorithm

Symbolic Analysis as Universal Tool for Deriving Properties of Non-linear Algorithms Case study of EM Algorithm Acta Polytechnica Hungarica Vol., No., 04 Sybolic Analysis as Universal Tool for Deriving Properties of Non-linear Algoriths Case study of EM Algorith Vladiir Mladenović, Miroslav Lutovac, Dana Porrat

More information

Projectile Motion with Air Resistance (Numerical Modeling, Euler s Method)

Projectile Motion with Air Resistance (Numerical Modeling, Euler s Method) Projectile Motion with Air Resistance (Nuerical Modeling, Euler s Method) Theory Euler s ethod is a siple way to approxiate the solution of ordinary differential equations (ode s) nuerically. Specifically,

More information

Enzyme kinetics: A note on negative reaction constants in Lineweaver-Burk plots

Enzyme kinetics: A note on negative reaction constants in Lineweaver-Burk plots Enzye kinetics: A note on negative reaction constants in Lineweaver-Burk plots Sharistha Dhatt# and Kaal Bhattacharyya* Departent of Cheistry, University of Calcutta, Kolkata 700 009, India #pcsdhatt@gail.co

More information

E0 370 Statistical Learning Theory Lecture 6 (Aug 30, 2011) Margin Analysis

E0 370 Statistical Learning Theory Lecture 6 (Aug 30, 2011) Margin Analysis E0 370 tatistical Learning Theory Lecture 6 (Aug 30, 20) Margin Analysis Lecturer: hivani Agarwal cribe: Narasihan R Introduction In the last few lectures we have seen how to obtain high confidence bounds

More information

1 Bounding the Margin

1 Bounding the Margin COS 511: Theoretical Machine Learning Lecturer: Rob Schapire Lecture #12 Scribe: Jian Min Si March 14, 2013 1 Bounding the Margin We are continuing the proof of a bound on the generalization error of AdaBoost

More information

Solutions of some selected problems of Homework 4

Solutions of some selected problems of Homework 4 Solutions of soe selected probles of Hoework 4 Sangchul Lee May 7, 2018 Proble 1 Let there be light A professor has two light bulbs in his garage. When both are burned out, they are replaced, and the next

More information

Research Article Integer-Valued Moving Average Models with Structural Changes

Research Article Integer-Valued Moving Average Models with Structural Changes Matheatical Probles in Engineering, Article ID 21592, 5 pages http://dx.doi.org/10.1155/2014/21592 Research Article Integer-Valued Moving Average Models with Structural Changes Kaizhi Yu, 1 Hong Zou, 2

More information

An Introduction to Meta-Analysis

An Introduction to Meta-Analysis An Introduction to Meta-Analysis Douglas G. Bonett University of California, Santa Cruz How to cite this work: Bonett, D.G. (2016) An Introduction to Meta-analysis. Retrieved fro http://people.ucsc.edu/~dgbonett/eta.htl

More information

A Better Algorithm For an Ancient Scheduling Problem. David R. Karger Steven J. Phillips Eric Torng. Department of Computer Science

A Better Algorithm For an Ancient Scheduling Problem. David R. Karger Steven J. Phillips Eric Torng. Department of Computer Science A Better Algorith For an Ancient Scheduling Proble David R. Karger Steven J. Phillips Eric Torng Departent of Coputer Science Stanford University Stanford, CA 9435-4 Abstract One of the oldest and siplest

More information

3D acoustic wave modeling with a time-space domain dispersion-relation-based Finite-difference scheme

3D acoustic wave modeling with a time-space domain dispersion-relation-based Finite-difference scheme P-8 3D acoustic wave odeling with a tie-space doain dispersion-relation-based Finite-difference schee Yang Liu * and rinal K. Sen State Key Laboratory of Petroleu Resource and Prospecting (China University

More information

Lost-Sales Problems with Stochastic Lead Times: Convexity Results for Base-Stock Policies

Lost-Sales Problems with Stochastic Lead Times: Convexity Results for Base-Stock Policies OPERATIONS RESEARCH Vol. 52, No. 5, Septeber October 2004, pp. 795 803 issn 0030-364X eissn 1526-5463 04 5205 0795 infors doi 10.1287/opre.1040.0130 2004 INFORMS TECHNICAL NOTE Lost-Sales Probles with

More information

A note on the multiplication of sparse matrices

A note on the multiplication of sparse matrices Cent. Eur. J. Cop. Sci. 41) 2014 1-11 DOI: 10.2478/s13537-014-0201-x Central European Journal of Coputer Science A note on the ultiplication of sparse atrices Research Article Keivan Borna 12, Sohrab Aboozarkhani

More information

HORIZONTAL MOTION WITH RESISTANCE

HORIZONTAL MOTION WITH RESISTANCE DOING PHYSICS WITH MATLAB MECHANICS HORIZONTAL MOTION WITH RESISTANCE Ian Cooper School of Physics, Uniersity of Sydney ian.cooper@sydney.edu.au DOWNLOAD DIRECTORY FOR MATLAB SCRIPTS ec_fr_b. This script

More information

Local asymptotic powers of nonparametric and semiparametric tests for fractional integration

Local asymptotic powers of nonparametric and semiparametric tests for fractional integration Stochastic Processes and their Applications 117 (2007) 251 261 www.elsevier.co/locate/spa Local asyptotic powers of nonparaetric and seiparaetric tests for fractional integration Xiaofeng Shao, Wei Biao

More information

A Low-Complexity Congestion Control and Scheduling Algorithm for Multihop Wireless Networks with Order-Optimal Per-Flow Delay

A Low-Complexity Congestion Control and Scheduling Algorithm for Multihop Wireless Networks with Order-Optimal Per-Flow Delay A Low-Coplexity Congestion Control and Scheduling Algorith for Multihop Wireless Networks with Order-Optial Per-Flow Delay Po-Kai Huang, Xiaojun Lin, and Chih-Chun Wang School of Electrical and Coputer

More information

Best Procedures For Sample-Free Item Analysis

Best Procedures For Sample-Free Item Analysis Best Procedures For Saple-Free Ite Analysis Benjain D. Wright University of Chicago Graha A. Douglas University of Western Australia Wright s (1969) widely used "unconditional" procedure for Rasch saple-free

More information

ANALYSIS OF HALL-EFFECT THRUSTERS AND ION ENGINES FOR EARTH-TO-MOON TRANSFER

ANALYSIS OF HALL-EFFECT THRUSTERS AND ION ENGINES FOR EARTH-TO-MOON TRANSFER IEPC 003-0034 ANALYSIS OF HALL-EFFECT THRUSTERS AND ION ENGINES FOR EARTH-TO-MOON TRANSFER A. Bober, M. Guelan Asher Space Research Institute, Technion-Israel Institute of Technology, 3000 Haifa, Israel

More information

The accelerated expansion of the universe is explained by quantum field theory.

The accelerated expansion of the universe is explained by quantum field theory. The accelerated expansion of the universe is explained by quantu field theory. Abstract. Forulas describing interactions, in fact, use the liiting speed of inforation transfer, and not the speed of light.

More information

TOURIST ARRIVALS AND ECONOMIC GROWTH IN SARAWAK

TOURIST ARRIVALS AND ECONOMIC GROWTH IN SARAWAK MPRA Munich Personal RePEc Archive TOURIST ARRIVALS AND ECONOMIC GROWTH IN SARAWAK Evan Lau and Swee-Ling Oh and Sing-Sing Hu Universiti Malaysia Sarawak, Universiti Malaysia Sarawak, Universiti Malaysia

More information

A MESHSIZE BOOSTING ALGORITHM IN KERNEL DENSITY ESTIMATION

A MESHSIZE BOOSTING ALGORITHM IN KERNEL DENSITY ESTIMATION A eshsize boosting algorith in kernel density estiation A MESHSIZE BOOSTING ALGORITHM IN KERNEL DENSITY ESTIMATION C.C. Ishiekwene, S.M. Ogbonwan and J.E. Osewenkhae Departent of Matheatics, University

More information

Ensemble Based on Data Envelopment Analysis

Ensemble Based on Data Envelopment Analysis Enseble Based on Data Envelopent Analysis So Young Sohn & Hong Choi Departent of Coputer Science & Industrial Systes Engineering, Yonsei University, Seoul, Korea Tel) 82-2-223-404, Fax) 82-2- 364-7807

More information

Sampling How Big a Sample?

Sampling How Big a Sample? C. G. G. Aitken, 1 Ph.D. Sapling How Big a Saple? REFERENCE: Aitken CGG. Sapling how big a saple? J Forensic Sci 1999;44(4):750 760. ABSTRACT: It is thought that, in a consignent of discrete units, a certain

More information

Detection and Estimation Theory

Detection and Estimation Theory ESE 54 Detection and Estiation Theory Joseph A. O Sullivan Sauel C. Sachs Professor Electronic Systes and Signals Research Laboratory Electrical and Systes Engineering Washington University 11 Urbauer

More information

Upper bound on false alarm rate for landmine detection and classification using syntactic pattern recognition

Upper bound on false alarm rate for landmine detection and classification using syntactic pattern recognition Upper bound on false alar rate for landine detection and classification using syntactic pattern recognition Ahed O. Nasif, Brian L. Mark, Kenneth J. Hintz, and Nathalia Peixoto Dept. of Electrical and

More information

Modified Systematic Sampling in the Presence of Linear Trend

Modified Systematic Sampling in the Presence of Linear Trend Modified Systeatic Sapling in the Presence of Linear Trend Zaheen Khan, and Javid Shabbir Keywords: Abstract A new systeatic sapling design called Modified Systeatic Sapling (MSS, proposed by ] is ore

More information

Moments of the product and ratio of two correlated chi-square variables

Moments of the product and ratio of two correlated chi-square variables Stat Papers 009 50:581 59 DOI 10.1007/s0036-007-0105-0 REGULAR ARTICLE Moents of the product and ratio of two correlated chi-square variables Anwar H. Joarder Received: June 006 / Revised: 8 October 007

More information

Bayesian Approach for Fatigue Life Prediction from Field Inspection

Bayesian Approach for Fatigue Life Prediction from Field Inspection Bayesian Approach for Fatigue Life Prediction fro Field Inspection Dawn An and Jooho Choi School of Aerospace & Mechanical Engineering, Korea Aerospace University, Goyang, Seoul, Korea Srira Pattabhiraan

More information

On Constant Power Water-filling

On Constant Power Water-filling On Constant Power Water-filling Wei Yu and John M. Cioffi Electrical Engineering Departent Stanford University, Stanford, CA94305, U.S.A. eails: {weiyu,cioffi}@stanford.edu Abstract This paper derives

More information

A Theoretical Analysis of a Warm Start Technique

A Theoretical Analysis of a Warm Start Technique A Theoretical Analysis of a War Start Technique Martin A. Zinkevich Yahoo! Labs 701 First Avenue Sunnyvale, CA Abstract Batch gradient descent looks at every data point for every step, which is wasteful

More information