On a new absolutely continuous bivariate generalized exponential distribution

Size: px
Start display at page:

Download "On a new absolutely continuous bivariate generalized exponential distribution"

Transcription

1 Statistical Methods and Applications anuscript No. (will be inserted by the editor On a new absolutely continuous bivariate generalized exponential distribution S. M. Mirhosseini M. Aini D. Kundu A. Dolati Received: date / Accepted: date Abstract In this paper we studied a three-paraeter absolutely continuous bivariate distribution whose arginals are generalized exponential distributions. The proposed three-paraeter bivariate distribution can be used quite effectively as an alternative to the Block and Basu bivariate exponential distribution. The oint probability density function, the oint cuulative distribution function and its associated copula have siple fors. We derive different properties of this new distribution. The axiu likelihood estiators of the unknown paraeters can be obtained by solving siultaneously three non-linear equations. We propose to use EM algorith to copute the axiu likelihood estiators, which can be ipleented quite conveniently. One data set has been analyzed for illustrative purposes. Finally we propose soe generalization of the proposed odel. Keywords Generalized exponential Bivariate exponential distribution Dependence Measure of association EM algorith Matheatics Subect Classification (2 MSC Priary 62E5 MSC Secondary 62H INTRODUCTION The two-paraeter generalized exponential (GE distribution proposed by Gupta and Kundu [] has been used quite successfully to analyze lifetie data and it has received soe attention in recent years; see, e.g.[3, 5, S. M. Mirhosseini Departent of Statistics, Faculty of Matheatical Science, Ferdowsi University of Mashhad,P.O. Box , Mashhad, Iran M. Aini Departent of Statistics, Faculty of Matheatical Science, Ferdowsi University of Mashhad,P.O. Box , Mashhad, Iran D. Kundu Departent of Matheatics and Statistics, Indian Institute of Technology Kanpur, Pin 286, India A. Dolati Departent of Statistics, Yazd University, Yazd, , Iran Tel.: Fax: E-ail: adolati@yazd.ac.ir

2 2 S. M. Mirhosseini et al 6]. A two-paraeter GE distribution with the shape paraeter α and the scale paraeter λ has the following cuulative distribution function (CDF ( F GE (x;α,λ = e λx α. ( Fro now on a GE with the CDF ( will be denoted by GE(α, λ. When α =, it becoes an exponential distribution with paraeter λ, and we will denote it by Exp(λ. It is well known that the probability density function (PDF and the hazard function (HF of a GE distribution can take different shapes. If the shape paraeter is less than or equal to one, the PDF and the HF are decreasing functions. When the shape paraeter is greater than one, then PDF is a uniodal, and the corresponding HF is an increasing function. For soe recent developents on the GE distribution, and for its different applications, the readers are referred to [, 8] and the references cited therein. In any reliability and life testing experients, bivariate data arise quite naturally. Aong the different bivariate lifetie odel, Marshall-Olkin bivariate exponential odel [22] is the ost popular one. It is a singular distribution, that is a pair (X,Y distributed as Marshall-Olkin odel has this property that P(X = Y >. Block and Basu [5] introduced a three-paraeter absolutely continuous bivariate exponential distribution by reoving the singular part of the Marshall-Olkin odel. Although Block and Basu bivariate exponential distribution does not have exponential arginals, but its arginals have decreasing PDFs and HFs. Several authors have been studied applications of the Block and Basu odel; see, e.g [2]. A wide survey on different bivariate distributions can be found in [4,7]. Recently, different bivariate exponential and bivariate generalized exponential (BGE distributions have been proposed and studied in the literature; see, e.g. [9 2, 24, 28, 32]. Most of these odels are singular distributions, and they can be used if there are ties in the data. It is not trivial to extend the univariate GE distribution to the bivariate or the ultivariate case. According to Joe [3], Section 4., a paraetric faily of distributions should satisfy four desirable properties: (a There should exist an interpretation like a ixture or other stochastic representation. (b The argins should belong to the sae paraetric faily and nuerical evaluation should be possible. (c The bivariate dependence between the argins should be described by a paraeter and cover a wide range of dependence. (d The distribution and density functions should preferably have a closed-for representation; at least nuerical evaluation should be possible. The ai of this paper is to study a new absolutely continuous bivariate generalized exponential distribution, whose arginals are univariate GE distributions and fulfill all of the properties (a (d. This new three-paraeter BGE distribution is obtained using the distribution of iniu order statistics of two independent saples of ordinary exponential distribution when the saple size is rando; see, e.g, [7]. All oint or arginal distributions and density functions, the corresponding oents, the copula and its density have siple analytic representations that can be easily eployed in applications. Moreover, the proposed distribution has positive quadrant dependent property which iplies that the easures of association such as Pearson s oent correlation, Kendall s tau and Spearan s rho for this distribution vary between and, which is a suitable range of dependence for applications. The axiu likelihood estiators (MLEs can be obtained by siultaneously solving three non-linear equations. The MLEs cannot be obtained in explicit fors. For known shape paraeter the MLEs of the scale

3 On a new absolutely continuous bivariate generalized exponential distribution 3 paraeters can be obtained by using the EM algorith. In the proposed EM algorith, at each E-step the axiization can be perfored explicitly. Hence the ipleentation of the EM algorith is quite siple in this case. Finally, the shape paraeter can be estiated by axiizing the profile log-likelihood function. The rest of the paper is organized as follows. We discuss the derivation of the BGE distribution in Section 2. In Section 3 we study its different properties. Statistical inference is provided in Section 4. The analysis of a real data set is presented in Section 5. Finally we propose soe generalizations and conclude the paper in Section 6. 2 BGE DISTRIBUTION: PDF AND CDF Following the idea given in [7], let {X,X 2,...} and {Y,Y 2,...} be two sequences of utually independent and identically distributed (i.i.d. rando variables. It is assued that for k {,2,3,...}, X k Exp(λ and Y k Exp(λ 2. Consider a sequence of independent Bernoulli trials in which the kth trial has probability of α k, < α, k {,2,3,...} and let N be the trial nuber on which the first success occurs. The discrete rando variable N has the probability ass function ( P(N = n = ( α α (... α α 2 n n ; n =,2,..., and its probability generating function is given by, see, e.g., [26] g(t = E(t N = ( t α ; t [,]. (2 Let U = in(x,...,x N and V = in(y,...,y N. The oint survival function of (U,V is then F(u,v = P{U u,v v} = n= [P(X i up(y i v] n P(N = n = g (e (λ u+λ 2 v = { e (λ u+λ 2 v } α. (3 Since F(u,v = F (u F 2 (v + F(u,v, the associated oint distribution function is given by F(u,v = ( e λu α + ( e λ2v α { e (λ u+λ 2 v } α, u,v,λ,λ 2 >, < α. (4 Fro (4, by taking v or u, it follows that U GE(α,λ and V GE(α,λ 2. (5 Therefore, it follows that the arginals of (4 are GE distributions. A pair (U,V distributed as (4 is said to have BGE with paraeters α, λ and λ 2, and it will be denoted by BGE(α,λ,λ 2. The oint density function of BGE is given by f (u,v = λ λ 2 αe (λ u+λ 2 v { αe (λ u+λ 2 v }{ e (λ u+λ 2 v } α 2. (6

4 4 S. M. Mirhosseini et al It can be easily seen that the oint PDF of BGE is a decreasing function in both arguents, and it resebles the oint PDF of the Block and Basu bivariate exponential distribution. Plots of the CDF and PDF of the BGE distribution are provided in Fig.. (a (b (c (d Fig. Plots of the oint CDF (left panels and PDF (right panels of BGE distribution. In (a, (b (α,λ,λ 2 = (.,, and in (c, (d (α,λ,λ 2 = (.9,,. Since < e (λ u+λ 2 v <, for u,v >, by using the binoial series expansion we have ( ( e (λ u+λ 2 α v α = ( e (λ u+λ 2 v, and the oint survival function (3 can be rewritten as ( α F(u,v = = = ( + e (λ u+λ 2 v.

5 On a new absolutely continuous bivariate generalized exponential distribution 5 We observe that the infinite series is suable, differentiable and hence by differentiating with respect to u and v we get f (u,v = λ λ 2 = ( α ( + 2 e (λ u+λ 2 v. (7 3 PROPERTIES 3. MOMENTS & CONDITIONAL MOMENTS First we provide expressions for the oint oent generating function (gf and Pearson s correlation coefficient of the BGE odel. Proposition If the rando vector (U,V BGE(λ,λ 2,α then (a the oint.g.f. of (U,V, for t < λ and s < λ 2, is given by ( α M U,V (t,s = ( + λ λ 2 2 ( λ t( λ 2 s, (8 = (b the Pearson s correlation coefficient of (U,V is given by ( α ( + Corr(U,V = 2 γ(α, 2, (9 κ(α, ( = where γ(α,β = Ψ(α + β Ψ(β, and κ(α,β = Ψ (β Ψ (α + β, is the digaa function, Ψ (. is its derivative and Ψ(t = d lnγ (t, dt is the gaa function [9]. Γ (t = x t e x dx, t >, Proof Starting fro M U,V (t,s = E(e tu+sv, using (7 we have ( α M U,V (t,s = λ λ 2 = ( + 2 e ( λ tx ( λ 2 sy dxdy. Since the quantity inside the suation is absolutely integrable, interchanging the suation and integration we have the required result. For part (b using (7, direct calculation shows that the product oent could be obtained as E(UV = Since U GE(α,λ and V GE(α,λ 2, fro [], using ( α ( + = λ λ 2 2. E(U = Ψ(α + Ψ( λ, E(V = Ψ(α + Ψ( λ 2

6 6 S. M. Mirhosseini et al and Var(U = Ψ ( Ψ (α + λ 2, Var(V = Ψ ( Ψ (α + λ2 2 the expression for correlation coefficient is iediate. The following result gives the conditional oents of BGE odel. Proposition 2 If the rando vector (U,V BGE(λ,λ 2,α, then the conditional expectation of V given U, is given by E(V U = u = λ λ 2 h U (u, where h U (u = f U (u/ F U (u is hazard rate function of U. Proof The conditional density function of (V U = u is given by ( α f V U=u (v = Thus, E(V U = u = = = = λ 2 αe λ u ( e λ u α x f X2 X =u(xdx α( e λ u α = αλ 2 ( e λ u α ( α = ( e λ u α αλ 2 e λ u ( e λ u α = λ λ 2 F U (u f U (u. = ( + 2 e ( λ u+λ 2 v. ( + e ( λ u ( α ( + e λ ( u ( λ 2 xe λ2x dx It is known that (see, e.g., [] the hazard rate function of univariate GE distribution with the shape paraeter α, has a decreasing hazard function if α. Therefore, we have the following result. Corollary If the rando vector (U,V BGE(λ,λ 2,α, then E(V U = u, is an increasing function in u. The following results will be useful for developent of the EM algorith. If (U,V and N are sae as defined before, then the oint PDF of U, V and N can be written as f (u,v,n = n 2 λ λ 2 e n(λ u+λ 2 v P(N = n; u >,v >,n =,2,... ( Hence the conditional probability of N = n, given U and V can be written as P(N = n U = u,v = v = Kn 2 e na P(N = n; n =,2,..., ( where A = λ u + λ 2 v, and K = αe A { αe A }{ e A } α 2.

7 On a new absolutely continuous bivariate generalized exponential distribution 7 Proposition 3 If (U,V BGE(λ,λ 2,α, and N is sae as defined before, then Proof If we denote E(N U = u,v = v = Kαe A ( e A α 3 { e 2A (α 2 6α e A (3α + 2 }. h(t = E ( te A N, then it follows that n 3 e A P(N = n = h ( + 3h ( + h (. n= Now the result follows by using the fact that fro (2 h(t = ( te A α ; t [,]. 3.2 STRESS-STRENGTH PARAMETER & DISTRIBUTION OF THE MINIMUM The stress-strength paraeter, R = P(U < V, is useful for data analysis purposes. The following result gives a convenient for for the stress-strength paraeter of BGE odel. Proposition 4 If (U,V BGE(λ,λ 2,α, then Proof Fro (7 we have P(U < V = = = y = = P(U < V = λ 2 λ + λ 2. f (x,ydxdy ( α y ( + 2 λ λ 2 e (λ x+λ 2 y dxdy ( α ( + λ 2 e λ2y ( e λy dy = λ 2 λ + λ 2 = ( α ( + which copletes the proof. = λ 2 λ + λ 2, Proposition 5 If (U,V BGE(λ,λ 2,α, then W = in(u,v GE(α,λ + λ 2. Proof P(W w = P(U w,v w which copletes the proof. = n= [P(X i wp(y i w] n P(N = n = g (e (λ +λ 2 w = { } e (λ α +λ 2 w

8 8 S. M. Mirhosseini et al 3.3 RÉNYI ENTROPY The entropy of a rando vector (U,V with the oint pdf f (x,y is a easure of variation of the uncertainty. Rényi entropy [27] is defined by for β >, where I R (β = log{t (β}, β T (β = Proposition 6 Suppose that (U,V BGE(λ,λ 2,α. Then f β (x,ydxdy. T (β = = k= ( β β(α 2 ( k α +β (λ λ 2 β ( + k + β 2. (2 Proof Fro (6 and substituting the transforations x = e λu and y = e λ2v, we have T (β = α β (λ λ 2 β (xy β ( β ( (α 2β αxy xy dxdy. By applying the binoial series expansion to the integrand and integrating with respect to x and y, we have the required result. 3.4 DEPENDENCE PROPERTIES In what follows we discuss the dependence properties of the BGE distribution through its associated copula. In view of Sklar s Theore [33], solving the equation C α {F (x,f 2 (y} = F(x,y, for the function C α : [,] [,] [,] yields the underlying copula associated with the pair (U,V having the BGE distribution defined by (4 as C α (u,v = u + v { ( u α ( v α } α = u + v uv{u α + v α } α, (3 for all u,v (, and < α. The theory and applications of copulas are well docuented in [25]. The survival copula Ĉ α (u,v = u + v +C α ( u, v, associated to C α is given by Ĉ α (u,v = ( u( v{( u α + ( v α } α, (4 for all u,v (, and < α. The copula (3 is not a eber of the Archiedean faily of copulas [25], but its associated survival copula (4 turns to be an Archiedean copula with the strict generator ϕ(t = ln( ( t α (see, Exaple 3 in [7]. Recall that for two copulas C and C 2, we say that C 2 is ore concordant than C (written C c C 2 if C (u,v C 2 (u,v, or equivalently Ĉ (u,v Ĉ 2 (u,v, for all u,v (,. A pair (X,X 2 with the copula C is positively quadrant dependent (written PQD if Π c C, where Π(u,v = uv is the product copula [25].

9 On a new absolutely continuous bivariate generalized exponential distribution 9 The dependence properties of the BGE distribution depend only on the paraeter α. The following result provides the dependence ordering of the BGE faily of distributions with respect to the paraeter α. Proposition 7 The copula C α defined by (3 is negatively ordered with respect to α; that is for α,α 2 (,], such that α α 2, we have C α2 c C α. Proof Since Ĉ α is an Archiedean copula with the generator ϕ α (t = ln( ( t α, and ϕ α2 ϕ α (t = ln( ( e t α 2 α is non-decreasing in t, for α α 2, in view of Theore in [25], we have Ĉ α2 c Ĉ α, or equivalently, C α2 c C α. Corollary 2 Suppose that (U,V BGE(λ,λ 2,α. Then (U,V is PQD. Proof As a consequence of Proposition 7, for α we have that Π(u,v = C (u,v C α (u,v for all u,v (,. Reark Since the BGE distribution defined by (4 has the PQD property, it is suitable to describe the positive dependence of a rando pair (U,V. However, it is very siple to consider a distribution to describe a negative dependence. It suffices to consider the copula C given by C α(u,v = u C α (u, v. It is obvious that the properties of a copula C α can be obtained in a siple way fro the corresponding properties of a copula C α ; see, [25] for detail. Let (U,V and (U,V be two continuous rando vectors with the sae univariate arginals and the respective oint density functions f and g. The pair (U,V is said to be ore positive likelihood ratio dependent (PLRD than the pair (U,V, denoted by (U,V PLRD (U,V, if f (x,y f (x 2,y 2 g(x,y 2 g(x 2,y f (x,y 2 f (x 2,y g(x,y g(x 2,y 2, (5 whenever x x 2 and x 2 y 2 [3]. When U and V are independent, then the pair (U,V is said to be PLRD and the condition (5 reduces to f (x,y f (x 2,y 2 f (x,y 2 f (x 2,y. Holland and Wang [2] showed that a sufficient condition for PLRD in the case of continuous rando variables is that 2 x y ln f (x,y. Proposition 8 Suppose that (U,V BGE(λ,λ 2,α. Then (U,V is PLRD. Proof Let (U,V be a rando vector with the oint distribution function G u (x,y = x α + y α (x + y xy α, x,y [,], < α. Let g u (.,. be the density function associated with G u (.,.. Since 2 x y lngu (x,y, for all x,y [,], < α, by the result of Holland and Wang [2] the pair (U,V is PLRD, i.e., (U,V PLRD (U,V, where U and V are independent with the sae univariate arginal distribution as U and V. For i =,2, let ϕ i (t be the increasing apping t λ i ln( t. Since the pair (U,V has the sae oint distribution as the pair (ϕ (U,ϕ 2 (V, in view of Theore 9.D.2 in [3], we have the required result. In the following we discuss the tail dependence properties of the BGE distribution. For a pair (U,V with the copula D, the lower (resp, upper tail dependence coefficient, λ L (resp, λ U is defined by [3,25] and λ L (U,V = li u + λ U (U,V = 2 li u D(u, u, (6 u D(u,u. (7 u

10 S. M. Mirhosseini et al Proposition 9 Suppose that (U,V BGE(λ,λ 2,α. Then λ L (U,V =, λ U (U,V =. Proof By taking into account (6, fro (3, the lower tail dependence coefficient of (U,V can be expressed as u(2 (2 u α α λ L (U,V = li u + =. u By taking into account (7, the upper tail dependence coefficient of (U,V can be expressed as u(2 (2 u α α λ U (U,V = 2 li u =. u In the rest of this section we provide expressions for soe well-known easures of association for a vector (U,V having BGE distribution. The population version of two of the ost coon nonparaetric easures of association between the coponents of a continuous rando pair (U,V are Kendall s tau (τ and Spearan s rho (ρ which depend only on the copula C of the pair (U,V, and are given by τ(c = 4 C(u,vdC(u,v, (8 and See [25] for detail. ρ(c = 2 C(u, vdudv 3. (9 Proposition Suppose that (U,V BGE(λ,λ 2,α. Then and τ(u,v = + 4αB(2,2α (Ψ(2 Ψ(2α +, ρ(u,v = 9 2α 2 =( ( α [B(,α] 2, where B(a,b = xa ( x b dx, denotes the beta function. Proof First recall that for a copula based easure of association κ, satisfying Scarsini s axios [3] and any pair of continuous rando variables (T, S with associated copula C, one has κ(t, S = κ( T, S, or equivalently, κ(c = κ(ĉ. Now the result follows fro Proposition 9 in [7]. Reark 2 Note that as a consequence of the PQD property of the BGE distribution, for (U,V BGE(λ,λ 2,α, we have Corr(U,V, τ(u,v and ρ(u,v. We provide the values of the Kendall s tau and the Spearan s rho for BGE distribution in Table. It is iediate that as α increases, both Kendall s tau and Spearan s rho decrease to, as it should be. Moreover, it is observed that ρ s(α τ(α = 3 as α, although we could not proved it theoretically. 2

11 On a new absolutely continuous bivariate generalized exponential distribution ρ α ρ(α τ(α s(α τ(α Table Kendall s tau and Spearan s rho associated with the faily of distributions (4 for different values of α (,]. 4 STATISTICAL INFERENCE 4. MAXIMUM LIKELIHOOD ESTIMATION In this section we discuss the MLEs of the paraeters of BGE distribution, based on a rando saple of size, naely {(u,v,...,(u,v } fro BGE(λ,λ 2,α. The log-likelihood function becoes ( ( l(θ = ln(αλ λ 2 v i + (α 2 ln e (λ u i +λ 2 v i + i= λ u i + λ 2 i= i= ( ln αe (λ u i +λ 2 v i where θ = (α,λ,λ 2. The axiu likelihood estiates can be obtained by axiizing (2 with respect to the unknown paraeters. The three noral equations becoe; l(θ α = ( α + ln e (λ u i +λ 2 v i e (λ x i +λ 2 y i i= i= αe (λ u i +λ 2 v i =, l(θ = u i e λ λ u i + (α 2 (λ u i +λ 2 v i i= i= e (λ u i +λ 2 v i + α u i e (λ u i +λ 2 v i i= αe (λ u i +λ 2 v i =, l(θ = v i e λ 2 λ 2 v i + (α 2 (λ u i +λ 2 v i i= e (λ u i +λ 2 v i + α v i e (λ u i +λ 2 v i i= αe (λ u i +λ 2 v i =. i= Note that the Newton-Raphson ethod or other optiization routine ay be used to axiize (2. To avoid that we propose to use the profile likelihood ethod to copute the MLEs of the unknown paraeters. For fixed α, the MLEs of λ and λ 2 can be obtained by axiizing the profile log-likelihood function with respect to λ and λ 2. We use EM algorith to copute the MLEs of λ and λ 2 for a given α, and finally we axiize the profile log-likelihood function of α, to copute the MLE of α. For ipleenting the EM algorith, we treat the proble as a issing value proble. Suppose along with (u,v, we observe the associated N value also. Therefore, the coplete observations are as follows: {(u,v,n,...,(u,v,n }. Based on the coplete saple, the log-likelihood function without the additive constant becoes i= (2 l c (λ,λ 2 = lnλ + lnλ 2 λ n i u i λ 2 i v i. (2 i= i=n

12 2 S. M. Mirhosseini et al For fixed α, the MLEs of λ and λ 2 becoe λ (α = i= n and λ 2 (α = iu i i= n. (22 iv i Now we are in a position to provide the EM algorith. For fixed α, at the E-step of the EM algorith, we construct the pseudo log-likelihood function at the k-th iterate by replacing the true value of N by its expected value. It takes the following for; l s (λ,λ 2 λ (k (k (α,λ 2 (α = lnλ + lnλ 2 λ u i E(N u i,v i,λ (k (k (α,λ 2 (α,α i= λ 2 i= v i E(N u i,v i,λ (k (k (α,λ 2 (α,α. (23 Here λ (k (k (α and λ 2 (α, are the values of λ and λ 2, respectively at the k-th iterate and E(N u i,v i,λ (k (k (α,λ 2 (α,α can be obtained using Proposition 4. Hence at the M-step, the axiization can be easily perfored to obtain λ (k+ (α and λ (k+ 2 (α as where C i = E(N u i,v i,λ (k λ (k+ (α = i= C iu i and λ (k+ 2 (α = i= C iv i, (24 (k (α,λ 2 (α,α. Continue the iteration until the convergence is et. Let us denote these estiates as λ (α and λ 2 (α. Finally axiize the profile log-likelihood function of α, to obtain the MLE of α, say α. Therefore, the MLEs of α, λ and λ 2 becoe α, λ ( α and λ 2 ( α, respectively. It can be easily seen that the density function of the BGE distribution satisfies all required conditions for the MLEs to be consistent and asyptotically norally distributed. We have the following result. Proposition If θ is the MLE of θ, then n(θ ˆθ d N 3 (,I. (25 Here d eans convergence in distribution and N 3 (,I, denotes the 3-variate noral distribution with ean vector and the covariance atrix I, and the atrix I is the Fisher inforation atrix; the eleents of the atrix I are presented in the Appendix. 4.2 TESTING OF HYPOTHESES We perfor the following two testing of hypotheses probles. PROBLEM : Testing whether the two arginals have the sae distributions or not, can be carried out as follows: H : λ = λ 2 versus H : λ λ 2. (26 In this case the MLE of λ = λ = λ 2 can be obtained along the sae line as before. The pseudo log-likelihood function becoes; l s (λ λ (k (α = 2lnλ λ i= (u i + v i E(N u i,v i,λ (k (α,α, (27

13 On a new absolutely continuous bivariate generalized exponential distribution 3 where λ (k denotes the estiate of λ at the k-th iteration. If D i = E(N u i,v i,λ (k (α, then λ (k+ (α = 2 i= D i(u i + v i. (28 Siilarly, as before the estiate of α can be obtained by axiizing the profile log-likelihood function of α. Now if we denote the estiates of α and λ as α and λ, respectively, then using the standard likelihood ratio, which has the asyptotic distribution as follows: ( 2 l( α, λ, λ 2 l( α, λ, λ χ 2. PROBLEM 2: If we want to test whether the two coponents are independent or not, the following test can be perfored: Under the null hypothesis the MLEs of λ and λ 2 can be obtained as H : α = versus H : α. (29 λ = i= u i and λ 2 = i= v i. (3 In this case, since α is in the boundary under H, the standard results do not apply. But using Theore 3 of Self and Liang [29], it follows that ( 2 l( α, λ, λ 2 l(, λ, λ χ2. (3 5 DATA ANALYSIS In this section we provide the analysis of a real data set fro McGilchrist and Aisbett [23]. The data has been obtained fro an experient, where M individuals are observed and ties between recurrence of a particular type of event are recorded. In this study the recurrence tie of infection in kidney patients who are using a portable dialysis achine are recorded. The infection occurs at the point of infection of the catheter, and when it occurs, the catheter has to be reoved, the infection cleared up and then the catheter reinserted. Recurrence ties are ties fro infection until next infection. For each patient two such recurrence ties are given; naely first recurrence tie (FRT and second recurrence tie (SRT. The data for 23 patients are reported in Table 2 Before, progressing further, we obtain soe basic statistics of the first recurrence tie and second recurrence tie, and they are reported in Table 3. It is clear fro Table 3 that both FRT and SRT have very long right tail. To get an idea about the shape of the epirical hazard function of the arginals, we provide the scaled TTT plots of FRT and SRT in Figure 2, as suggested by Aarset []. This plot provides an idea of the shape of the hazard function of a distribution. It has been shown in [] that the scaled TTT transfor is convex (concave if the hazard rate is decreasing (increasing. For this data set, it indicates that both of variables have decreasing epirical hazard functions. Note that fro Table 3 the Spearan s rho and Kendall s tau for FRT and SRT data are given by.266 and.84, respectively. This observation deonstrates an obvious positive dependence between involved data. This observation shows that the proposed BGE ay be used for analyzing this bivariate data set.

14 4 S. M. Mirhosseini et al No. FRT SRT No. FRT SRT Table 2 Kidney infection data of 23 patients. The patient No., first recurrence tie (FRT, second recurrence tie (SRT are reported. Statistics FRT SRT Miniu 2 8 st Quartile 3 26 Median 34 4 Mean st Quartile Maxiu Standard deviation Pearson s corr..9 Kendall s tau.84 Spearan s rho.266 rho/tau.449 Table 3 Descriptive statistics of the data vector. We divide all the data points by ainly for coputational purposes, it is not going to affect in the statistical inference. To get initial estiates of λ and λ 2, we fit exponential distribution to the arginals, and obtain the initial estiates of λ and λ 2 as.93 and.86. For each α, we use these initial estiates to start the EM algorith. The profile log-likelihood function of α becoes a uniodal function, and it is reported in Figure 3. Finally axiizing the profile log-likelihood function of α, we obtain the MLEs of α, λ and λ 2, and they are α =.799, λ =.7559 and λ 2 =.655, and the corresponding log-likelihood value is The associated 95% confidence intervals are.799.6, and , respectively. To see whether the GE distribution fits the arginal data or not, we copute the Kologorov- Sirnov (KS distances of GE(.799,.7559 and GE(.799,.655 to the epirical CDF of FRT and SRT data, respectively. It is observed that the KS distance between GE(.799,.7559 and epirical CDF of FRT is.73 and the associated p value is Siilarly, the KS distance between GE(.799,.655 and epirical CDF of SRT is.689 and the associated p value is Therefore, the GE distribution can be used to fit the arginals reasonably well, for the above data set.

15 On a new absolutely continuous bivariate generalized exponential distribution (a (b Fig. 2 Scaled TTT plots of (a FRT and (b SRT. 49 Profile log likelihood function α Fig. 3 Profile log-likelihood function of α. We perfor the following two testing of hypotheses probles; naely (26 and (29. In the first case (26, the value of the test statistic is.22. Since the p =.6466, we do not reect the null hypothesis. In the second case the value of test statistic is In this case p =.74, hence we reect the null hypothesis. Note that fro Table we see that the population versions of Kendall s tau and Spearan s rho for BGE distribution with estiated paraeter α =.7 are given by.285 and.94, respectively. These values are close to the Spearan s rho and Kendall s tau between FRT and SRT, given in Table 3. This observation also shows that the proposed BGE ay be used for analyzing this bivariate data set. The natural question that arises here is whether the BGE odel fits these bivariate data or not. In the next section we provide a copula goodness-of-fit test.

16 6 S. M. Mirhosseini et al For coparison purposes we have fitted (i the three paraeters Block and Basu [5] bivariate exponential odel with the pdf f (x,y = λ λ(λ 2 +λ 3 λ +λ 2 e λ x (λ 2 +λ 3 y, x < y λ 2 λ(λ +λ 3 λ +λ 2 e (λ +λ 3 x λ 2 y, x > y where λ,λ 2,λ 3 > and λ = λ +λ 2 +λ 3 and (ii the five-paraeter absolutely continuous bivariate generalized exponential distribution of the for F(x,y = [( e λ x α + ( e λ 2x α 2 ] α, which is constructed based on the Clayton s copula [25] proposed in Kundu and Gupta [2], to this data set. (32 Model Estiated paraeters Log-likelihood BEG ˆα =.799 ˆλ =.7559 ˆλ2 = Block and Basu odel ˆλ =.9297 ˆλ2 =.8586 ˆλ3 = Kundu and Gupta odel ˆα =.562 ˆα =.46 ˆλ =.68 ˆα 2 =.5392 ˆλ2 = Table 4 The MLEs and the values of Loglikelihood The MLEs and corresponding log-likelihood values are given in Table 4. Therefore, based on the loglikelihood values, we can say that the proposed BGE odel provides a better fit than Block and Basu [5] bivariate exponential odel and is coparable with the Kundu and Gupta [2] odel for this data set. 5. A COPULA GOODNESS-OF-FIT TEST Once a odel has been stated and estiated the natural question is to check whether the initial odel assuptions are realistic. In other words we are faced with the so-called goodness-of-fit proble. As an advantage of the Sklar s Theore [25] the arginal distributions and the copula can be chosen independently of one another. The univariate GE distribution provide adequate descriptions of the FRT and SRT data, individually. Since the copula of the BGE odel has a closed and siple for, one can also try to perfor a copula goodness-of-fit test. A review and coparison of goodness-of-fit procedures is given in [8]. Let (x,y,...,(x n,y n be observations fro a rando vector (X,Y. When dealing with bivariate data, the ost natural way of checking the adequacy of a copula odel would be to copare the fitted copula and the epirical copula (see, e.g, [6] of data defined by C n ( i n, n = nuber of pairs (x,y in the saple with x x (i,y y (, n where x (i and y (, i, n, denote order statistics fro the saple. Figure 4 shows graph of the pairs (x i,y i and (u i,v i, i =,...,23, for FRT and SRT data. Fro siulations provided in [8] a good cobination of power and conceptual siplicity is provided by the Craer-von Mises statistic: S n = n i= (C α (u i,v i C n (u i,v i 2,

17 On a new absolutely continuous bivariate generalized exponential distribution 7 where C α is the fitted copula and u i = rank of x i aong x,...,x n n + and v i = rank ofy i aong y,...,y n. n + This statistic easures how close the fitted copula is fro the epirical copula of data. The P-value of the test is coputed using a paraetric bootstrap procedure described in Appendix A of [8]. To this end, we applied this procedure to both the FRT and SRT data. The bootstrap values S,...,S of the Craer-von Mises test statistic are generated and we found the proportion of these values that are larger than S n =.394 as P-value.3. Thus we ay conclude that the copula C α defined by (3 with the association paraeter α =.799 perfors a good fit for this data set. SRT noralized ranks (SRT FRT noralized ranks (FRT (a (b Fig. 4 Pairs of observations (a and pairs of noralised ranks (b for FRT and SRT data. 6 CONCLUSIONS In this paper we studied a bivariate absolutely continuous generalized exponential distribution, whose arginals are generalized exponential distributions. It has three paraeters and the arginals have decreasing hazard functions. Therefore, the proposed odel can be used as an alternative to the quite popular Block and Basu bivariate exponential odel. We derive different properties of the proposed odel, and also provide the EM algorith for coputation of the MLEs of the unknown paraeters. We have analyzed one real data set, and it is observed that the proposed odel provides a good fit to the data set. Now we briefly discuss different generalizations of the proposed odel. ( Although we have developed the ethodology for the bivariate case, along the sae line ultivariate generalized exponential distribution also can be defined. Several properties can be obtained even for the ultivariate case also. (2 Instead of taking

18 8 S. M. Mirhosseini et al exponential distribution for X and Y, it is possible to develop the ethodology when they follow Weibull distribution. In this case we can obtain bivariate/ ultivariate exponentiated Weibull distribution as an alternative to the already-existent bivariate Weibull odels; see, e.g [4]. It will be interesting to develop different properties of this ore flexible distributions. More work is needed along these directions. ACKNOWLEDGEMENTS: The authors would like to thank the referees for their helpful coents which iproved significantly the earlier draft of the paper. APPENDIX. FISHER INFORMATION MATRIX Fro (2 we have 2 l(θ α 2 2 l(θ λ 2 2 l(θ λ 2 2 = n n [ α 2 e (λ ] x i +λ 2 y i 2 i= αe (λ, x i +λ 2 y i = n λ 2 = n λ 2 2 (α 2 (α 2 n i= n i= x 2 i e (λ x i +λ 2 y i x 2 i e (λ x i +λ 2 y i n ( e (λ x i +λ 2 y i 2 α i= ( αe (λ x i +λ 2 y i 2, y 2 i e (λ x i +λ 2 y i y 2 i e (λ x i +λ 2 y i n ( e (λ x i +λ 2 y i 2 α i= ( αe (λ x i +λ 2 y i 2, n i= 2 l(θ x i e (λ x i +λ 2 y i n = α λ ( e (λ x i +λ 2 y i 2 + x i e (λ x i +λ 2 y i i= ( αe (λ x i +λ 2 y i 2, 2 n l(θ y i e = α λ 2 (λ x i +λ 2 y i n i= ( e (λ x i +λ 2 y i 2 + y i e (λ x i +λ 2 y i i= ( αe (λ x i +λ 2 y i 2, 2 n l(θ x i y i e = (α 2 λ λ 2 (λ x i +λ 2 y i n ( e (λ x i +λ 2 y i 2 α x i y i e (λ x i +λ 2 y i i= ( αe (λ x i +λ 2 y i 2. i= The Fisher inforation is I(θ = [I(θ i ], where I i (θ = E 2 l(θ θ i θ, and θ = (θ,θ 2,θ 3 = (α,λ,λ 2. We shall now present the exact expressions of I i (θ, for i =,2,3. Direct calculations show that ( 2 l(θ I = E α 2 ( 2 l(θ I 22 = E λ 2 = n α 2 = n λ 2 ( + = k= ( k α +3( α 2 k ( + k ( + 2α(α 2A + αa 2, ( 2 l(θ I 33 = E λ2 2 = n λ2 2 ( + 2α(α 2A + αa 2 ( 2 l(θ I 2 = E α λ ( 2 l(θ I 3 = E α λ 2 ( 2 l(θ I 23 = E λ λ 2 = nα = nα λ (B + B 2, (B + B 2, λ 2 = nα ((α 2C + αc 2, λ λ 2,

19 On a new absolutely continuous bivariate generalized exponential distribution 9 where A = A 2 = ( ( α 4 ( ( α ( + 3 4, ( α 2 ( α k (k , = = k= B = B 2 = C = C 2 = = ( ( α 3 ( + ( +( α 2 α k ( + k + 2 3, = k= = ( ( α 4 ( ( ( α 2 α k ( + k = k= ( α ( ( α ( + 3 4,, References. Aarset, M.V. (987. How to identify a bathtub hazard rate?, IEEE Transactions on Reliability, 36, Achcar, J. A., Coelho-Barros, E. A. and Mazucheli, J. (23. Block and Basu bivariate lifetie distribution in the presence of cure fraction, Journal of Applied Statistics, 4, Asgharzadeh, A. (28. Approxiate MLE for the scaled generalized exponential distribution under progressive type-ii censoring, Journal of the Korean Statistical Society, 38, Balakrishnan, N. and Lai, C. D. (29. Continuous Bivariate Distributions, second edition. Springer, New York. 5. Block, H. and Basu, A. P. (974. A continuous bivariate exponential extension, Journal of the Aerican Statistical Association, 69, Deheuvels, P. (979. La fonction de dépendance epirique et ses propriétés Un test non paraétrique díndépendance, Acadéie Royale de Belgique Bulletin de la Classe des Sciences 5e Série, 65, Dolati, A., Aini, M. and Mirhosseini, S. M. (24. Dependence properties of bivariate distributions with proportional (reversed hazards arginals, Metrika, 77, Genest, C., Reillard, B. and Beaudoin, D. (28. Goodness-of-fit tests for copulas: A review and a power study. Insurance: Matheatics and Econoics, 42, Gradshteyn, I. S. and Ryzhik, I. M. (2. Table of Integrals, Series and Products, sixth ed., Acadeic Press, San Diego.. Gupta, R. D. and Kundu, D. (999. Generalized exponential distributions, Austral. New Zealand J. Statist, 4, Gupta, R. D. and Kundu, D. (27. Generalized exponential distribution; existing ethods and soe recent developents, J. Statistical Planning and Inference, 37, Holland, P. W. and Wang, Y. J. (987. Dependence functions for continous bivariates densities, Coun. Statist. Theory and Methods, 6:3, Joe, H. (997. Multivariate Models and Dependence Concepts, Chapan & Hall, London. 4. Jose, K. K., M., Ristíc, M. M. and Joseph, A. (2. Marshall Olkin bivariate Weibull distributions and processes, Statistical Papers, 52, Kannan, N., Kundu, D, Nair, P. and Tripathi, R. C. (2. The generalized exponential cure rate odel with covariates, Journal of Applied Statistics, 37:, Ki, N. (24. Approxiate MLE for the scale paraeter of the generalized exponential distribution under rando censoring, Journal of the Korean Statistical Society, 43:, 9 3.

20 2 S. M. Mirhosseini et al 7. Kotz, S., Balakrishnan, N. and Johnson, N.L. (2. Continuous ultivariate distributions, Vol., second ed., Wiley: New York. 8. Kundu, D. and Gupta, R. D. (28. Generalized exponential distribution: Bayesian estiations, Coputational Statistics and Data Analysis, 52, Kundu, D. and Gupta, R. D. (29. Bivariate generalized exponential distribution, J. Multivariate Analysis,, Kundu, D. and Gupta, R. D. (2. Modified Sarhan Balakrishnan singular bivariate distribution, J. Statistical Planning and Inference, 4, Kundu, D. and Gupta, R. D. (2. Absolute continuous bivariate generalized exponential distribution, AStA Adv. Stat. Anal., 95, Marshall, A. W. and Olkin, I. (967. A ultivariate exponential distribution, Journal of the Aerican Statistical Association, 62, McGilchrist, C. A. and Aisbett, C. W. (99. Regression with frailty in survival analysis, Bioetrics, 47, Mohsin, M., Kazianka, H., Pilz, J. and Gebhardt, A. (24. A new bivariate exponential distribution for odeling oderately negative dependence, Stat. Methods Appl, 23, Nelsen, R. B. (26. An Introduction to Copulas. Second Edition, Springer, New York. 26. Pillai, R. N. and Jayakuar, K. (995. Discrete Mittag Leffler distributions, Statistics and Probability Letters, 23, Rényi, A. (96. On easures of entropy and inforation, in: Proceedings of the 4th Berkeley Syposiu on Matheatical Statistics and Probability, vol. I, University of California Press, Berkeley, pp Sarhan, A. and Balakrishnan, N. (27. A new class of bivariate distributions and its ixture, Journal of Multivariate Analysis, 98, Self, S.G. and Linag, K. L. (987. Asyptotic properties of the axiu likelihood estiators, and likelihood ratio test under nonstandard conditions, Journal of the Aerican Statistical Association, 82, Scarsini, S. (984. On easures of concordance. Stochastica, 8, Shaked, M. and Shanthikuar, G. (27. Stochastic Orders, Springer, New York. 32. Shoaee, S. and Khorra, E. (22. A new absolute continuous bivariate generalized exponential distribution, J. Statistical Planning and Inference, 42, Sklar, A. (959. Fonctions de répartition à n diensions et leurs arges. Publ. Inst. Statist. Univ. Paris 8,

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

On Sarhan-Balakrishnan Bivariate Distribution

On Sarhan-Balakrishnan Bivariate Distribution J. Stat. Appl. Pro. 1, No. 3, 163-17 (212) 163 Journal of Statistics Applications & Probability An International Journal c 212 NSP On Sarhan-Balakrishnan Bivariate Distribution D. Kundu 1, A. Sarhan 2

More information

BIVARIATE NONCENTRAL DISTRIBUTIONS: AN APPROACH VIA THE COMPOUNDING METHOD

BIVARIATE NONCENTRAL DISTRIBUTIONS: AN APPROACH VIA THE COMPOUNDING METHOD South African Statist J 06 50, 03 03 BIVARIATE NONCENTRAL DISTRIBUTIONS: AN APPROACH VIA THE COMPOUNDING METHOD Johan Ferreira Departent of Statistics, Faculty of Natural and Agricultural Sciences, University

More information

Classical and Bayesian Inference for an Extension of the Exponential Distribution under Progressive Type-II Censored Data with Binomial Removals

Classical and Bayesian Inference for an Extension of the Exponential Distribution under Progressive Type-II Censored Data with Binomial Removals J. Stat. Appl. Pro. Lett. 1, No. 3, 75-86 (2014) 75 Journal of Statistics Applications & Probability Letters An International Journal http://dx.doi.org/10.12785/jsapl/010304 Classical and Bayesian Inference

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

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

Bivariate Geometric (Maximum) Generalized Exponential Distribution

Bivariate Geometric (Maximum) Generalized Exponential Distribution Bivariate Geometric (Maximum) Generalized Exponential Distribution Debasis Kundu 1 Abstract In this paper we propose a new five parameter bivariate distribution obtained by taking geometric maximum of

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

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

Best Linear Unbiased and Invariant Reconstructors for the Past Records

Best Linear Unbiased and Invariant Reconstructors for the Past Records BULLETIN of the MALAYSIAN MATHEMATICAL SCIENCES SOCIETY http:/athusy/bulletin Bull Malays Math Sci Soc (2) 37(4) (2014), 1017 1028 Best Linear Unbiased and Invariant Reconstructors for the Past Records

More information

MODELING DEPENDENT RISKS WITH MULTIVARIATE ERLANG MIXTURES ABSTRACT KEYWORDS

MODELING DEPENDENT RISKS WITH MULTIVARIATE ERLANG MIXTURES ABSTRACT KEYWORDS MODELING DEPENDENT RISKS WITH MULTIVARIATE ERLANG MIXTURES BY SIMON C.K. LEE AND X. SHELDON LIN ABSTRACT In this paper, we introduce a class of ultivariate Erlang ixtures and present its desirable properties.

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

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

Inferences using type-ii progressively censored data with binomial removals

Inferences using type-ii progressively censored data with binomial removals Arab. J. Math. (215 4:127 139 DOI 1.17/s465-15-127-8 Arabian Journal of Matheatics Ahed A. Solian Ahed H. Abd Ellah Nasser A. Abou-Elheggag Rashad M. El-Sagheer Inferences using type-ii progressively censored

More information

Weighted Marshall-Olkin Bivariate Exponential Distribution

Weighted Marshall-Olkin Bivariate Exponential Distribution Weighted Marshall-Olkin Bivariate Exponential Distribution Ahad Jamalizadeh & Debasis Kundu Abstract Recently Gupta and Kundu [9] introduced a new class of weighted exponential distributions, and it can

More information

Using EM To Estimate A Probablity Density With A Mixture Of Gaussians

Using EM To Estimate A Probablity Density With A Mixture Of Gaussians Using EM To Estiate A Probablity Density With A Mixture Of Gaussians Aaron A. D Souza adsouza@usc.edu Introduction The proble we are trying to address in this note is siple. Given a set of data points

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

Discriminating Between the Bivariate Generalized Exponential and Bivariate Weibull Distributions

Discriminating Between the Bivariate Generalized Exponential and Bivariate Weibull Distributions Discriminating Between the Bivariate Generalized Exponential and Bivariate Weibull Distributions Arabin Kumar Dey & Debasis Kundu Abstract Recently Kundu and Gupta ( Bivariate generalized exponential distribution,

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

Estimating Parameters for a Gaussian pdf

Estimating Parameters for a Gaussian pdf Pattern Recognition and achine Learning Jaes L. Crowley ENSIAG 3 IS First Seester 00/0 Lesson 5 7 Noveber 00 Contents Estiating Paraeters for a Gaussian pdf Notation... The Pattern Recognition Proble...3

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

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

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

This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and

This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and This article appeared in a ournal published by Elsevier. The attached copy is furnished to the author for internal non-coercial research and education use, including for instruction at the authors institution

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

are equal to zero, where, q = p 1. For each gene j, the pairwise null and alternative hypotheses are,

are equal to zero, where, q = p 1. For each gene j, the pairwise null and alternative hypotheses are, Page of 8 Suppleentary Materials: A ultiple testing procedure for ulti-diensional pairwise coparisons with application to gene expression studies Anjana Grandhi, Wenge Guo, Shyaal D. Peddada S Notations

More information

Bayes Estimation of the Logistic Distribution Parameters Based on Progressive Sampling

Bayes Estimation of the Logistic Distribution Parameters Based on Progressive Sampling Appl. Math. Inf. Sci. 0, No. 6, 93-30 06) 93 Applied Matheatics & Inforation Sciences An International Journal http://dx.doi.org/0.8576/ais/0063 Bayes Estiation of the Logistic Distribution Paraeters Based

More information

Support recovery in compressed sensing: An estimation theoretic approach

Support recovery in compressed sensing: An estimation theoretic approach Support recovery in copressed sensing: An estiation theoretic approach Ain Karbasi, Ali Horati, Soheil Mohajer, Martin Vetterli School of Coputer and Counication Sciences École Polytechnique Fédérale de

More information

Kernel Methods and Support Vector Machines

Kernel Methods and Support Vector Machines Intelligent Systes: Reasoning and Recognition Jaes L. Crowley ENSIAG 2 / osig 1 Second Seester 2012/2013 Lesson 20 2 ay 2013 Kernel ethods and Support Vector achines Contents Kernel Functions...2 Quadratic

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

arxiv: v1 [stat.ot] 7 Jul 2010

arxiv: v1 [stat.ot] 7 Jul 2010 Hotelling s test for highly correlated data P. Bubeliny e-ail: bubeliny@karlin.ff.cuni.cz Charles University, Faculty of Matheatics and Physics, KPMS, Sokolovska 83, Prague, Czech Republic, 8675. arxiv:007.094v

More information

Scale Invariant Conditional Dependence Measures

Scale Invariant Conditional Dependence Measures Sashank J. Reddi sjakkar@cs.cu.edu Machine Learning Departent, School of Coputer Science, Carnegie Mellon University Barnabás Póczos bapoczos@cs.cu.edu Machine Learning Departent, School of Coputer Science,

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

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

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

Fairness via priority scheduling

Fairness via priority scheduling Fairness via priority scheduling Veeraruna Kavitha, N Heachandra and Debayan Das IEOR, IIT Bobay, Mubai, 400076, India vavitha,nh,debayan}@iitbacin Abstract In the context of ulti-agent resource allocation

More information

Intelligent Systems: Reasoning and Recognition. Perceptrons and Support Vector Machines

Intelligent Systems: Reasoning and Recognition. Perceptrons and Support Vector Machines Intelligent Systes: Reasoning and Recognition Jaes L. Crowley osig 1 Winter Seester 2018 Lesson 6 27 February 2018 Outline Perceptrons and Support Vector achines Notation...2 Linear odels...3 Lines, Planes

More information

Two Weighted Distributions Generated by Exponential Distribution

Two Weighted Distributions Generated by Exponential Distribution Journal of Mathematical Extension Vol. 9, No. 1, (2015), 1-12 ISSN: 1735-8299 URL: http://www.ijmex.com Two Weighted Distributions Generated by Exponential Distribution A. Mahdavi Vali-e-Asr University

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

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

Research Article On the Isolated Vertices and Connectivity in Random Intersection Graphs

Research Article On the Isolated Vertices and Connectivity in Random Intersection Graphs International Cobinatorics Volue 2011, Article ID 872703, 9 pages doi:10.1155/2011/872703 Research Article On the Isolated Vertices and Connectivity in Rando Intersection Graphs Yilun Shang Institute for

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

On Weighted Exponential Distribution and its Length Biased Version

On Weighted Exponential Distribution and its Length Biased Version On Weighted Exponential Distribution and its Length Biased Version Suchismita Das 1 and Debasis Kundu 2 Abstract In this paper we consider the weighted exponential distribution proposed by Gupta and Kundu

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

Estimation of the Population Mean Based on Extremes Ranked Set Sampling

Estimation of the Population Mean Based on Extremes Ranked Set Sampling Aerican Journal of Matheatics Statistics 05, 5(: 3-3 DOI: 0.593/j.ajs.05050.05 Estiation of the Population Mean Based on Extrees Ranked Set Sapling B. S. Biradar,*, Santosha C. D. Departent of Studies

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

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

Empirical phi-divergence test-statistics for the equality of means of two populations

Empirical phi-divergence test-statistics for the equality of means of two populations Epirical phi-divergence test-statistics for the equality of eans of two populations N. Balakrishnan, N. Martín and L. Pardo 3 Departent of Matheatics and Statistics, McMaster University, Hailton, Canada

More information

On Pearson families of distributions and its applications

On Pearson families of distributions and its applications Vol. 6(5) pp. 08-7 May 03 DOI: 0.5897/AJMCSR03.0465 ISSN 006-973 03 Acadeic Journals http://www.acadeicournals.org/ajmcsr African Journal of Matheatics and Coputer Science Research Full Length Research

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

Identical Maximum Likelihood State Estimation Based on Incremental Finite Mixture Model in PHD Filter

Identical Maximum Likelihood State Estimation Based on Incremental Finite Mixture Model in PHD Filter Identical Maxiu Lielihood State Estiation Based on Increental Finite Mixture Model in PHD Filter Gang Wu Eail: xjtuwugang@gail.co Jing Liu Eail: elelj20080730@ail.xjtu.edu.cn Chongzhao Han Eail: czhan@ail.xjtu.edu.cn

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

Bivariate Sinh-Normal Distribution and A Related Model

Bivariate Sinh-Normal Distribution and A Related Model Bivariate Sinh-Normal Distribution and A Related Model Debasis Kundu 1 Abstract Sinh-normal distribution is a symmetric distribution with three parameters. In this paper we introduce bivariate sinh-normal

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

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

Constructing Locally Best Invariant Tests of the Linear Regression Model Using the Density Function of a Maximal Invariant

Constructing Locally Best Invariant Tests of the Linear Regression Model Using the Density Function of a Maximal Invariant Aerican Journal of Matheatics and Statistics 03, 3(): 45-5 DOI: 0.593/j.ajs.03030.07 Constructing Locally Best Invariant Tests of the Linear Regression Model Using the Density Function of a Maxial Invariant

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

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

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

Closed-form evaluations of Fibonacci Lucas reciprocal sums with three factors

Closed-form evaluations of Fibonacci Lucas reciprocal sums with three factors Notes on Nuber Theory Discrete Matheatics Print ISSN 30-32 Online ISSN 2367-827 Vol. 23 207 No. 2 04 6 Closed-for evaluations of Fibonacci Lucas reciprocal sus with three factors Robert Frontczak Lesbank

More information

Uniform Approximation and Bernstein Polynomials with Coefficients in the Unit Interval

Uniform Approximation and Bernstein Polynomials with Coefficients in the Unit Interval Unifor Approxiation and Bernstein Polynoials with Coefficients in the Unit Interval Weiang Qian and Marc D. Riedel Electrical and Coputer Engineering, University of Minnesota 200 Union St. S.E. Minneapolis,

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

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

Explicit solution of the polynomial least-squares approximation problem on Chebyshev extrema nodes

Explicit solution of the polynomial least-squares approximation problem on Chebyshev extrema nodes Explicit solution of the polynoial least-squares approxiation proble on Chebyshev extrea nodes Alfredo Eisinberg, Giuseppe Fedele Dipartiento di Elettronica Inforatica e Sisteistica, Università degli Studi

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

DERIVING PROPER UNIFORM PRIORS FOR REGRESSION COEFFICIENTS

DERIVING PROPER UNIFORM PRIORS FOR REGRESSION COEFFICIENTS DERIVING PROPER UNIFORM PRIORS FOR REGRESSION COEFFICIENTS N. van Erp and P. van Gelder Structural Hydraulic and Probabilistic Design, TU Delft Delft, The Netherlands Abstract. In probles of odel coparison

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

Generalized Exponential Geometric Extreme Distribution

Generalized Exponential Geometric Extreme Distribution Generalized Exponential Geometric Extreme Distribution Miroslav M. Ristić & Debasis Kundu Abstract Recently Louzada et al. ( The exponentiated exponential-geometric distribution: a distribution with decreasing,

More information

Note on generating all subsets of a finite set with disjoint unions

Note on generating all subsets of a finite set with disjoint unions Note on generating all subsets of a finite set with disjoint unions David Ellis e-ail: dce27@ca.ac.uk Subitted: Dec 2, 2008; Accepted: May 12, 2009; Published: May 20, 2009 Matheatics Subject Classification:

More information

Generalized AOR Method for Solving System of Linear Equations. Davod Khojasteh Salkuyeh. Department of Mathematics, University of Mohaghegh Ardabili,

Generalized AOR Method for Solving System of Linear Equations. Davod Khojasteh Salkuyeh. Department of Mathematics, University of Mohaghegh Ardabili, Australian Journal of Basic and Applied Sciences, 5(3): 35-358, 20 ISSN 99-878 Generalized AOR Method for Solving Syste of Linear Equations Davod Khojasteh Salkuyeh Departent of Matheatics, University

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

Estimation of P[Y X] for generalized Pareto distribution

Estimation of P[Y X] for generalized Pareto distribution Fro the SelectedWorks of anijeh ahoodi Winter February, Estiation of P[Y X] for generalized Pareto distribution anijeh ahoodi sadegh rezaei rasool tahasbi Available at: https://works.bepress.co/anijeh_ahoodi//

More information

Assessment of wind-induced structural fatigue based on joint probability density function of wind speed and direction

Assessment of wind-induced structural fatigue based on joint probability density function of wind speed and direction The 1 World Congress on Advances in Civil, Environental, and Materials Research (ACEM 1) eoul, Korea, August 6-3, 1 Assessent of wind-induced structural fatigue based on oint probability density function

More information

On the Comparison of Fisher Information of the Weibull and GE Distributions

On the Comparison of Fisher Information of the Weibull and GE Distributions On the Comparison of Fisher Information of the Weibull and GE Distributions Rameshwar D. Gupta Debasis Kundu Abstract In this paper we consider the Fisher information matrices of the generalized exponential

More information

E. Alpaydın AERFAISS

E. Alpaydın AERFAISS E. Alpaydın AERFAISS 00 Introduction Questions: Is the error rate of y classifier less than %? Is k-nn ore accurate than MLP? Does having PCA before iprove accuracy? Which kernel leads to highest accuracy

More information

Statistics and Probability Letters

Statistics and Probability Letters Statistics and Probability Letters 79 2009 223 233 Contents lists available at ScienceDirect Statistics and Probability Letters journal hoepage: www.elsevier.co/locate/stapro A CLT for a one-diensional

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

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

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

Burr Type X Distribution: Revisited

Burr Type X Distribution: Revisited Burr Type X Distribution: Revisited Mohammad Z. Raqab 1 Debasis Kundu Abstract In this paper, we consider the two-parameter Burr-Type X distribution. We observe several interesting properties of this distribution.

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

Boosting with log-loss

Boosting with log-loss Boosting with log-loss Marco Cusuano-Towner Septeber 2, 202 The proble Suppose we have data exaples {x i, y i ) i =... } for a two-class proble with y i {, }. Let F x) be the predictor function with the

More information

Bipartite subgraphs and the smallest eigenvalue

Bipartite subgraphs and the smallest eigenvalue Bipartite subgraphs and the sallest eigenvalue Noga Alon Benny Sudaov Abstract Two results dealing with the relation between the sallest eigenvalue of a graph and its bipartite subgraphs are obtained.

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

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

arxiv: v1 [math.pr] 17 May 2009

arxiv: v1 [math.pr] 17 May 2009 A strong law of large nubers for artingale arrays Yves F. Atchadé arxiv:0905.2761v1 [ath.pr] 17 May 2009 March 2009 Abstract: We prove a artingale triangular array generalization of the Chow-Birnbau- Marshall

More information

ASSUME a source over an alphabet size m, from which a sequence of n independent samples are drawn. The classical

ASSUME a source over an alphabet size m, from which a sequence of n independent samples are drawn. The classical IEEE TRANSACTIONS ON INFORMATION THEORY Large Alphabet Source Coding using Independent Coponent Analysis Aichai Painsky, Meber, IEEE, Saharon Rosset and Meir Feder, Fellow, IEEE arxiv:67.7v [cs.it] Jul

More information

Compression and Predictive Distributions for Large Alphabet i.i.d and Markov models

Compression and Predictive Distributions for Large Alphabet i.i.d and Markov models 2014 IEEE International Syposiu on Inforation Theory Copression and Predictive Distributions for Large Alphabet i.i.d and Markov odels Xiao Yang Departent of Statistics Yale University New Haven, CT, 06511

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

Curious Bounds for Floor Function Sums

Curious Bounds for Floor Function Sums 1 47 6 11 Journal of Integer Sequences, Vol. 1 (018), Article 18.1.8 Curious Bounds for Floor Function Sus Thotsaporn Thanatipanonda and Elaine Wong 1 Science Division Mahidol University International

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

Bayes Decision Rule and Naïve Bayes Classifier

Bayes Decision Rule and Naïve Bayes Classifier Bayes Decision Rule and Naïve Bayes Classifier Le Song Machine Learning I CSE 6740, Fall 2013 Gaussian Mixture odel A density odel p(x) ay be ulti-odal: odel it as a ixture of uni-odal distributions (e.g.

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

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

Univariate and Bivariate Geometric Discrete Generalized Exponential Distributions

Univariate and Bivariate Geometric Discrete Generalized Exponential Distributions Univariate and Bivariate Geometric Discrete Generalized Exponential Distributions Debasis Kundu 1 & Vahid Nekoukhou 2 Abstract Marshall and Olkin (1997, Biometrika, 84, 641-652) introduced a very powerful

More information

On Bivariate Birnbaum-Saunders Distribution

On Bivariate Birnbaum-Saunders Distribution On Bivariate Birnbaum-Saunders Distribution Debasis Kundu 1 & Ramesh C. Gupta Abstract Univariate Birnbaum-Saunders distribution has been used quite effectively to analyze positively skewed lifetime data.

More information

. The univariate situation. It is well-known for a long tie that denoinators of Pade approxiants can be considered as orthogonal polynoials with respe

. The univariate situation. It is well-known for a long tie that denoinators of Pade approxiants can be considered as orthogonal polynoials with respe PROPERTIES OF MULTIVARIATE HOMOGENEOUS ORTHOGONAL POLYNOMIALS Brahi Benouahane y Annie Cuyt? Keywords Abstract It is well-known that the denoinators of Pade approxiants can be considered as orthogonal

More information

Ştefan ŞTEFĂNESCU * is the minimum global value for the function h (x)

Ştefan ŞTEFĂNESCU * is the minimum global value for the function h (x) 7Applying Nelder Mead s Optiization Algorith APPLYING NELDER MEAD S OPTIMIZATION ALGORITHM FOR MULTIPLE GLOBAL MINIMA Abstract Ştefan ŞTEFĂNESCU * The iterative deterinistic optiization ethod could not

More information

Can the Threshold Performance of Maximum Likelihood DOA Estimation be Improved by Tools from Random Matrix Theory?

Can the Threshold Performance of Maximum Likelihood DOA Estimation be Improved by Tools from Random Matrix Theory? Advances in Signal Processing 2(): 8-28, 204 DOI: 0.389/asp.204.02003 http://www.hrpub.org Can the Threshold Perforance of axiu Likelihood DOA Estiation be Iproved by Tools fro Rando atrix Theory? Yuri

More information

SPECTRUM sensing is a core concept of cognitive radio

SPECTRUM sensing is a core concept of cognitive radio World Acadey of Science, Engineering and Technology International Journal of Electronics and Counication Engineering Vol:6, o:2, 202 Efficient Detection Using Sequential Probability Ratio Test in Mobile

More information