On Perturbed Symmetric Distributions Associated with the Truncated Bivariate Elliptical Models
|
|
- Felicity Walsh
- 5 years ago
- Views:
Transcription
1 Communications of the Korean Statistical Society Vol. 15, No. 4, 8, pp On Perturbed Symmetric Distributions Associated with the Truncated Bivariate Elliptical Models Hea-Jung Kim 1) Abstract This paper proposes a class of perturbed symmetric distributions associated with the bivariate elliptically symmetricor simply bivariate elliptical) distributions. The class is obtained from the nontruncated marginals of the truncated bivariate elliptical distributions. This family of distributions strictly includes some univariate symmetric distributions, but with extra parameters to regulate the perturbation of the symmetry. The moment generating function of a random variable with the distribution is obtained and some properties of the distribution are also studied. These developments are followed by practical examples. Keywords: Perturbed symmetric distribution; truncated bivariate elliptical distribution; Skew-elliptical distribution. 1. Introduction A distribution of k 1 random vector X, written X EC k θ, Σ, g k) ), is said to have k-variate elliptically symmetricor simply elliptical) distribution with location vector θ R k and a k k positive definite) dispersion matrix Σ and the density generating function g k). The density of X distribution is given by fx θ, Σ) = Σ 1 g k) [ x θ) Σ 1 x θ) ], 1.1) for some density generator function g k) u), u, such that ) k Γ u k 1 g k) u)du =. 1.) By varying the function g k), distributions with longer or shorter tails than the k-variate normal can be obtained. The density 1.1) has appeared at various places in the literature, sometimes in a somewhat casual manner. A systematic treatment of this distribution has been given by Fang et al. 199) and Fang and Zhang 199). Various univariate/multivariate extensions of the distribution is considered by Azzalini 1985), This work was partially supported by the Korea Research Foundation Grant funded by the Korean Government MOEHRD) KRF C16). 1) Professor, Department of Statistics, Dongguk University, 3 Pil-Dong, Chung-Gu, Seoul 1-715, Korea. kim3hj@dongguk.edu π k
2 484 Hea-Jung Kim Branco and Dey 1), Azzalini and Capitanio 3), Ma and Genton 4), Arellano- Valle et al. 6) and Kim 7, 8) among others. The major goal of this paper is to study a class of perturbed symmetric distributions as an extension of univariate elliptically symmetric distribution. The class is obtained from deriving the distribution of Z [X a < Y < b], 1.3) where X, Y ) is a bivariate elliptical variable with the location parameter θ 1, θ ) and scale matrix Σ. Although cases of perturbed distributionsobtained from a conditioning mechanism) are well addressed in the literature, so far as we know, there are few results concerning the class of perturbed distributions associated with a doubly truncated bivariate elliptical distribution. This motivates the investigation in this article. The interest in studying the distribution 1.3) comes from both theoretical and applied directions. On the theoretical side, it provides a class of distributions that enjoys a number of formal properties which resemble those of the elliptical distributions as given in Fang and Zhang 199) and produces a general class of distributions that strictly includes many symmetric distributions as well as the univariate skewed distributions considered by Arellano-Valle et al. 6). From the applied viewpoint, the distribution is a unimodal empirical distribution with presence of skewness and possibly heavyor light) tail see, Figure 3.1). This implies that 1.3) is useful to modeling random phenomena which have heavieror lighter) tails than the normal as well as having some skewness. Moreover, the class of distributions obtained from 1.3) provides yet other models that enable us to analyze a screened data in terms of the sum of truncated and untruncated observations.. Preliminaries Prior to suggesting the class of perturbed symmetric distributions, we provide lemmas useful for studying property of the class. If we set Ψ = {ψ ij } with ψ 11 = ψ = 1 and ψ 1 = ψ 1 = ρ, the properties of EC, Ψ, g ) ) distribution yield following theorems. From now on, we will use Ψ to denote this standard form of dispersion matrix. Lemma.1 Let W EC, Ψ, g ) ), W = W 1, W ). Then the conditional distribution of W given that W 1 = w is EC 1 α, β, g qw) ) distribution. Here α = ρw, β = 1 ρ, g qw) = g ) u + qw))/g 1) qw)), where g 1) u) = g ) r + u)dr and qw) = w. Proof: Straightforward application of the result of Branco and Dey 1, pp. 1) yields the result. Lemma. Let W EC, Ψ, g ) ), W = W 1, W ). If Z is set to equal to W 1 conditionally on a < W < b. Then the pdf of Z is f Z z) = f g 1)z) { F gqz) λ 1 b λz) F gqz) λ 1 a λz) }, for z R,.1) F g 1)b) F g 1)a) where λ = ρ/ 1 ρ, λ 1 = 1 + λ ) 1/ = 1/ 1 ρ and f g 1) ) and F g 1) ) are the pdf and the cdf of EC 1, 1, g 1) ), respectively. F gqz) is the cdf of EC 1, 1, g qz) ) with qz) = z.
3 Perturbed Distribution Associated with Truncated Elliptical Model 485 Proof: The pdf of Z can be expressed as f Z z) = P a < W < b z)f W1 z). P a < W < b) Using the property of EC, Ψ, g ) ) see, Fang et al. 199, p.43) we obtain the marginal distribution, W 1 EC 1, 1, g 1) ). Further Lemma.1 gives W W 1 = z EC 1 α, β, g qz) ) and hence P a < W < b z) = a α P < W α < b α ) z β β β = F gqz) λ 1 b λz) F gqz) λ 1 a λz). Noticing that f W1 z) = f g 1)z) and P a < W < b) = F g 1)b) F g 1)a), we have the result. 3. Perturbed Symmetric Distribution A generalization of Lemma. yields the following result. Let X = X, Y ) and let X EC θ, Σ, g ) ) with θ = θ 1, θ ), Σ = {σ ij }, σ ii = σi and σ 1 = ρσ 1 σ. Then, the pdf of Z [X a < Y < b] variable is fzz) = f g 1)u 1z)) { F λ gq z) 1ub) λu 1 z)) F λ gq z) 1ua) λu 1 z)) } { σ 1 Fg1) ub)) F g1) ua)) }, 3.1) for z R, where u 1 z) = z θ 1 )/σ 1, ua) = a θ )/σ, ub) = b θ )/σ, F gq z) is the cdf of EC 1, 1, g q z)) with q z) = u 1 z). The density of Z variable is obtained from.1) using the transformation relations X = σ 1 W 1 + θ 1 and Y = σ W + θ. From now on, we use the same notations without redefining. Definition 3.1 If a random variable Z has density function 3.1), then we say that Z is a perturbed symmetric random variable with parameters θ, Σ and a non-negative perturb function { Fgq z) λ 1ub) λu 1 z)) F λ gq z) 1ua) λu 1 z)) } pz; θ, σ, ρ) = For brevity we shall also say that Z is PS a,b) θ, Σ, g ) ). { Fg1) ub)) F. 3.) g 1) ua))} We see, from 3.1), that the perturbed distribution arise when the symmetric density f g 1)u 1 z))/σ 1 of potential observation z gets distorted so that it is multiplied by some non-negative perturb function pz; θ, σ, ρ). From Lemma.1 and 3.1), we can get an alternative and convenient expression for the pdf of PS a,b) θ, Σ, g ) ) distribution as f Zz) = ub) [ {r g ) ρu1 z)} ] ua) 1 ρ + u 1 z) dr σ { }, z R. 3.3) 1 1 ρ F g1) ub)) F g1) ua))
4 486 Hea-Jung Kim There are many subclasses of bivariate elliptical distributions, such as bivariate normals, bivariate t-distributions, symmetric bivariate Pearson types VII, bivariate scale mixture normal distributions and some others see, Fang et al. 199, for the others). These are the most useful subclasses from both practical and theoretical aspects. In the following examples, we see that 3.3) yields some perturbed symmetric distributions from these subclasses. Example 3.1 A Perturbed Normal Distribution) The generator function for a bivariate normal is g ) N u) = π) 1 exp u ). 3.4) Thus the density fz z) of Z PS a,b)θ, Σ, g ) N ) variable is obtained from 3.3) and it is fzz) = φu 1z)) {Φλ 1 ub) λu 1 z)) Φλ 1 ua) λu 1 z))}, z R. 3.5) σ 1 {Φub)) Φua))} This agrees with the density given in Kim 7). When a =, b =, θ = and Σ = Ψ, we see that 3.5) reduces to the density of skew-normal distribution SN ) by Azzalini 1985). Example 3. A Perturbed t ν Distribution) The generator function for a bivariate t-distribution with the degrees of freedom ν is g ) ν u) = C 1 ν + u) ν+, 3.6) where C 1 = ν ν/ Γ{ν + )/}/π Γ{ν/}). It follows from 3.3) that the pdf fz z) of PS a,b) θ, Σ, g ν ) ) distribution is fzz) = f νu 1 z)){f ν+1 v z)) F ν+1 v 1 z))}, z R, 3.7) σ 1 {F ν ub)) F ν ua))} where f ν ) and F ν ) denote the pdf and the cdf of a univariate standard t ν distribution, while F ν+1 ) is the cdf of a univariate standard t ν+1 distribution. Example 3.3 A Perturbed Pearson Type VII Distribution) A 1 random vector X is said to have a symmetric bivariate Pearson Type VII distribution if it has a density generator function g ) V II u) = C 1 + u m) N, N > 1, m > 3.8) and C = πm) 1 ΓN)/ΓN 1). Some analytic algebra plugging 3.8) in 3.3) leads to the pdf of PS a,b) θ, Σ, g ) V II ) distribution given by f Zz) = δσ 1 ) 1 f N 1,m u 1 z)){f N,m+1 w z)) F N,m+1 w 1 z))}, z R, 3.1) where δ = F N 1/,m ub)) F N 1/,m ua)), w 1 z) = {λ 1ua) λu 1 z)} m + 1 m + u1 z) and w z) = {λ 1ub) λu 1 z)} m + 1 m + u1 z).
5 Perturbed Distribution Associated with Truncated Elliptical Model 487 Here f N 1/,m ) and F N 1/,m ) denote the pdf and the cdf of a univariate standard Pearson type VII distribution with parameters N 1/ and m, while F N,m+1 ) is the cdf of a univariate standard Pearson type VII distribution with parameters N and m +1. The perturbed Pearson type VII distribution includes a number of important distributions such as the perturbed t distributionfor N=m/+1) and the perturbed Cauchy distributionfor m=1 and N=3/). Example 3.4 A Perturbed Scale Mixture of Normal Distribution) The generator function for a bivariate scale mixture of normal is { g ) H,K u) = {πkη)} 1 exp u } dhη), 3.11) Kη) where η is a mixing variable with the cdf Hη) and Kη) is a weight function. From 3.3), the pdf fz z) of the perturbed scale mixture of normal distribution, written PS a,b) θ, Σ, g ) H,K ), is given as 1 πδ 1 Kη)σ 1 1 ρ ub) ua) exp [ {r ρu 1z)} Kη)1 ρ ) + u 1z) ] r Hη). Kη) Transforming {r ρu 1 z)}/ Kη)1 ρ ) to w yields ) fzz) 1 u = 1 z) φ {Φu 1z)) Φu z))}dhη), 3.1) δ 1 σ 1 Kη) Kη) for z R, where δ 1 = ) ) ub) ua) Φ Φ dhη), Kη) Kη) u 1z) = λ 1ub) λu 1 z) Kη) and u z) = λ 1ua) λu 1 z) Kη). Note that 3.1) reduces to 3.5) when Hη) is degenerate with Kη) = 1. Also note that 3.1) is reduced to 3.7) when Kη) = 1/η with Hη) as the cdf of a gamma distribution, i.e., η Gν/, /ν) so that νη χ ν. Other combinations of Hη) and Kη) functions for the scale mixture of normal distribution are considered by Chen and Dey 1998) and Branco and Dey 1), among other. Those combinations can be applied to 3.1) to coming at various classes of perturbed symmetric distributions. For example, through the scale mixtures by Chen and Dey 1998), a class of perturbed stable densities, perturbed logistic densities and that of perturbed symmetric power densities can also be obtained from 3.1). 4. Property of the Distribution 4.1. Distributional properties Now we will state some interesting properties for the PS a,b) θ, Σ, g ) ) distribution as well as the associate examples. Some properties are focused on those of PS a,b) θ, Σ, g ) K,H )
6 488 Hea-Jung Kim Figure 3.1: Various Shapes of perturbed symmetric about zero densities with ρ =.8 a) the pdf of SN λ) PE, ), Ψ, g ) N ) distribution, where λ = ρ/ 1 ρ ; b) the pdf of PE.5,), Ψ, g ) N ) distribution; c) the pdf of PE.5,), Ψ, g ν ) ) distribution with ν = 3; d) the pdf of PE.5,), Ψ, g ) V II ) distribution with m = N = 7) which incudes several perturbed distributions obtained from well-known symmetric distributions. Property 4.1. If Z PS a,b) θ, Σ, g ) ), then Z PS a,b) θ, Σ, g ) ), where θ and Σ are obtained from θ and Σ by changing their elements θ 1 and ρ with θ 1 and ρ, respectively. Property 4.. Let UZ) = Z θ 1 )/σ 1, where Z PS a,b) θ, Σ, g ) ). Then Property 4.3. If Z PS θ, )θ, Σ, g ) ), its pdf is UZ) PS ua),ub)), Ψ, g ) ). 4.1) f Z z) = σ 1 f g 1)u 1 z))f gq z) λu 1 z)), for z R, 4.) where F gq z) ) is the cdf of EC 1, 1, g q z)) with q z) = u 1 z). The density 4.) does not depend on the values of θ and σ and it is equivalent to the univariate skew-elliptical distribution introduced by Branco and Dey 1). In
7 Perturbed Distribution Associated with Truncated Elliptical Model 489 this sense, Definition 3.1 extends the result of Branco and Dey 1) where they studied only for the case of θ =, σ = 1) and a =, b = ). When the relationship between λ and ρ in 4.) is considered, we see λ has the sign of ρ, ranges from to and is equal to zero if and only if ρ =. This directly gives the following property. Property 4.4. Suppose X EC θ, Σ, g ) ), where X = X, Y ), θ = θ 1, θ ), Σ = {σ ij } and σ 1 = ρσ 1 σ. If ρ = then [X Y > θ ] [X Y θ ] EC 1 θ 1, σ 11, g 1) ). 4.3) One can give a probabilistic representation of the distribution PS a,b) θ, Σ, g ) H,K ) in terms of a scale mixture of independent normal and truncated normal laws. The scale mixture scheme directly gives the following theorem. Theorem 4.1. Conditionally on η having the cdf Hη), let V Nµ 1, Kη)τ1 ) and W Nµ, Kη)τ ) be independent random variables. Then, for any real values a 1 and a with a 1, Z = a 1 V + a W Ia,b) PS a,b) θ, Σ, g ) H,K ), 4.4) where W Ia,b) denotes a truncated Nµ, Kη)τ ) variable and a and b are lower and upper truncation points, respectively. θ = θ 1, θ ) with θ 1 = i=1 a iµ i and θ = µ ; Σ = {σ ij }, σ ii = σi, with σ 11 = Kη) i=1 a i τ i, σ = Kη)τ and ρ = a σ /σ 1. Proof: Let U = a 1 V + a W. Then U, W ) is a bivariate normal variable with N θ, Σ) distribution. Conditionally on η, the distribution of U conditionally on a < W < b is PS a,b) θ, Σ, g ) N ), by Example 1. Given η, V and W in U = a 1V + a W are independent normal variables and hence the distribution of U given that a < W < b equals that of a 1 V + a W Ia,b) figuring in the statement of Theorem 4.1. Thus the scale mixture distribution 4.4) is directly obtained by using Example 4. When a =, Z is a scale mixture of normal distribution. Thus the representation in 4.4) reveals the structure of the class of PS a,b) θ, Σ, g ) H,K ) distributions and indicates the kind of departure from the symmetric distribution. Furthermore, the representation provides one-for-one method of generating a random variable Z with density 3.1). For generating the truncated normal variable W Ia,b), the one-for-one method by Devroye 1986) may be used. Corollary 4.1 Conditionally on η, let V and W be independent N, Kη)) random variables and let X, Y ) N, Kη)Ψ), where η is a random variable with the cdf Hη) and weight function Kη). Then 1 V + λ W 1 + λ 1 + λ Ia,b) [X a < Y < b] PS a,b), Ψ, g ) H,K ), 4.5) where λ = ρ/ 1 ρ. Proof: Setting µ 1 = µ = and τ 1 = τ = 1, we have the result from Theorem 4.1.
8 49 Hea-Jung Kim For Kη) = 1, we see that 4.5) is equivalent to the probabilistic representation by Kim 7). In the case a = and b =, if we set Kη) = 1 and Kη) = 1/η, PS a,b), Ψ, g ) H,K ) distributions reduce to SN λ) by Azzalini 1985) and the skew-t ν distribution, respectively. Their probabilistic representations obtained from 4.5) agree with those of SN λ) given in Henze 1986) and skew-t ν derived by Kim ). 4.. Moments of the perturbed distributions In this section we derive a general expression for the moment generating functionmgf) for the perturbed scale mixture of normal distribution in 3.1). To compute the moments of Z PS a,b) θ, Σ, g ) H,K ), it suffices to compute the moments of UZ) = Z θ 1)/σ 1. We see from 3.1) that UZ) has the density 1 guz)) = δ 1 Kη) ) uz) φ {Φu 1z)) Φu z))}dhη), 4.6) Kη) for z R. Some algebra gives that the moment generating function of UZ) = Z θ 1 )/σ 1 is M UZ) t) = 1 δ 1 E η [{Φ u t b)) Φ u t a))}e Kη) t ], 4.7) for t R, where E η denotes that the expectation is taken with respect to the distribution of η variate having the density dhη)/dη and u t a) = ua) ρ Kη) t, u t b) = ub) ρ Kη) t. Kη) Kη) Naturally, the moments of UZ) = Z θ 1 )/σ 1 can be obtained by using the moment generating function differentiation. For example: E[UZ)] = M UZ) t) t= = ρ [ ] E η Kη) {φu b)) φu a))}, 4.8) δ 1 where u a) = ua)/ Kη) and u b) = ub)/ Kη). Unfortunately, for higher moments this rapidly becomes tedious. An alternative procedure makes use of the fact that d dx [xk+1 φx)] = k + 1)x k φx) x k+ φx), 4.9) for k = 1,, 1,, 3,..., yields the following result. Let W = UZ)/ Kη). Under the distribution 4.6), the relation 4.9) and integrating by parts gives E[k + 1)W k W k+ ], that is ρ δ 1 λ k+1 1 [ φu b))e [V + λu b)] k+1 φu a))e [V + λu a)] k+1] dhη), 4.1) for k = 1,, 1,..., where V is a N, 1) variate.
9 Perturbed Distribution Associated with Truncated Elliptical Model 491 By setting k = 1,, 1, we obtain three expressions, which may be solved to yield the first three moments of UZ). Higher moments could be found similarly. One obtains, E[UZ)] = ρ [ ] E η Kη){φu b)) φu a))}, δ 1 [ ] E[UZ) ] =E η [Kη)] ρ E η Kη){u b)φu b)) u a)φu a))}, δ 1 E[UZ) 3 ] = ρ E η [Kη) 3 {3 ρ + u b)ρ )φu b)) δ 1 ] 3 ρ + u a)ρ )φu a))}. By using the Binomial expansion, one can see that the general formula for the moments of Z PE a,b) θ, Σ, g ) H,K ) is E[Z k ] = k j= ) k θ k j 1 σ j 1 j E[UZ)j ]. 4.11) When θ 1 =, E[Z k ] = σ1 k E[UZ) k ]. It is also noted that existence of the moments depends on the mixing distribution Hη). The class of perturbed scale mixture of normal distribution includes well-known skewelliptical distributions as in the following two examples. EXAMPLE 5. One member of the perturbed scale mixture normal distribution in 3.1) is the PS θ, )θ, Σ, g ) H,K ) distribution, for which H is degenerate with Kη) = 1. From 3.1), we obtain the pdf given by f Zz) = σ 1 φuz))φλu 1 z)), z R. 4.1) This distribution is equivalent to the skew-normal distribution, SN θ 1, σ 1, λ), by Azzalini 1985). From 4.11), one finds the moments, ) E[Z] = θ 1 + σ 1 ρ π, VarZ) = σ 1 1 ρ π and α 3 = ) 4 π 1 π ρ3 ) 3 1 ρ. 4.13) π These values of Z agree with those given in Arnold et al. 1993). See Azzalini 1985) and Henze 1986) for the other properties of the distribution. EXAMPLE 6. Another member of the perturbed scale mixture of normal distribution distribution is the skew-t ν distribution. When Kη) = 1/η and η Gν/, /ν), the pdf 3.1) of the PS θ, )θ, Σ, g ) H,K ) distribution reduces to fzz) = λu 1 z) ) ν + 1 f ν u 1 z))f ν+1, z R, 4.14) σ 1 ν + u1 z)
10 49 Hea-Jung Kim where f ν ) and F ν ) denote the pdf and the cdf of a univariate standard t ν distribution, while F ν+1 ) is the cdf of a univariate standard t ν+1 distribution. This distribution is equivalent to the skew-t ν distribution, ST θ 1, σ 1, ν, λ), by Kim ). From 4.11), we obtain, ν EZ) = θ 1 + σ 1 ρ π VarZ) = σ1 ν Γ {ν 1)/}, for ν > 1, Γ {ν/} ν EZ), for ν >, A α 3 =, VarZ) 3 for ν > 3, 4.15) where [ ν ) 1 )] A = σ1λ 3 3νΓ{ν 1)/} 1 π 1 + λ ) 3 ν )Γ{ν/} ν 3 + λ B, [ ] ν ) ν )Γ{ν 1)/} ρ B = + 3ν 3) 3πΓ{ν/} 1 and λ =. 1 ρ It is easily seen that the distribution does not depend on θ and σ and leads to a parametric class of distributions that have strict inclusion of t ν distribution for the case θ 1 =, σ 1 = 1 and ρ = ) and perturbed Cauchy distributionfor ν = 1). See Kim ) for the other properties and applications of the distribution 4.14). Note that B > by using a numerical evaluation for ν > 3. Thus 4.13) and 4.15) imply that the distributions in both examples are skewed to the rightthe left) when ρ > ρ < ). 5. Applications The aim of this section is to provide simple applications of the material presented so far, focusing on those of PS a,b) θ, Σ, g ) N ) and PS a,b)θ, Σ, g ν ) ) distributions Comparing elasticities in regression Suppose we have a model where quantity Q depends on price P, so that the demand function is Q = αp β u. 5.1) When we take logs of 5.1), we obtain log Q = log α + β log P + log u which is of the standard form Y = α + βx + e. 5.) In the econometric context, β is called the price elasticity of the demand Q. The price elasticity is formally defined as the relative change in quantity demanded divided by the relative change in price, i.e. β=percent change in Q)/percent change in P ). See Rudy ) for the analysis of the price elasticity to diagnose consumer expenditure. Let observational models of two quantitiesq 1 and Q ) in 5.1) be y ij = α i + β i x ij + e ij, i = 1, ; j = 1,..., n i, 5.3)
11 Perturbed Distribution Associated with Truncated Elliptical Model 493 iid where e ij N, τi ) and e 1j and e j are independent variables. Suppose, from the microeconomic theory, that the price elasticity of the demand for Q 1 is inelastic to price so that β 1, 1). Then the price elasticities of the demands for Q 1 and Q can be compared by the distribution of β /β 1 obtained from the Bayesian approach. We shall take a reference prior that is independently uniform in αi, β i and log τi, so that παi, β i, τi ) 1/τ i I < β 1 < 1). It is now clear that the marginal posterior distributions of β i s i = 1, ) are independent and β 1 Data T N b 1, τ 1 ) I < β 1 < 1) and β Data N b, τ ) S 11 S for known τi, i = 1,, β 1 Data T t ni ) b 1, s 1 S 11 ) I < β 1 < 1) and β Data t ni ) b, for unknown τi, where b i = S i /S ii, n i )s i = S Sy/S ii, S ii = n i j=1 x ij x i ), S i = n i j=1 y ij ȳ i )x ij x i ), S = n i j=1 y ij ȳ i ), x i = n i j=1 x ij/n i and ȳ i = ni j=1 y ij/n i. Here T Na, b)i < β 1 < 1) and T t ni 1)a, b)i < β 1 < 1) denote the normal and t ni 1) distributions truncated below at and above at 1, respectively. a and b are the location and scale parameters of each distribution. For known τ i, the posterior distribution function F c) of β /β 1 is the same as P Z ) obtained from Z PS,1) θ c, Σ c, g ) N ) by Theorem 4.1, where θ c = θ 1, θ ) with θ 1 = b cb 1 and θ = b 1 ; Σ c = {σ ij } with σ 11 = c τ1 /S 11 + τ /S, σ = τ1 /S 11, σ 1 = cτ1 /S 11 and ρ = σ 1 / σ 11 σ. Thus the distribution function Fc z) of PS,1) θ c, Σ c, g ) N ) variable in 3.5) leads to the distribution function F c) of β /β 1. That is F c) = F c ) = 1 L θ 1, θ ), ρ σ11 σ ) 1 θ Φ Φ σ L θ 1 σ11, 1 θ σ, ρ ) θ σ ), where Lα, β, ρ) denotes the orthant probability function of the standard bivariate normal variable defined in Johnson and Kotz 197). For the case τi is unknown but with large n i i = 1, ), an approximate posterior distribution function of β /β 1 can be obtained by substituting n i s i /n i ) for τi, i = 1, in the scale matrix Σ c of the PS,1) θ c, Σ c, g ) N ) distribution. 5.. Paired sample problem Let X 1, Y 1 ),..., X n, Y n ) be paired random sample from bivariate normal distribution, where CorrX i, Y i ) = ρ, X i Nµ X, σ iid iid ) and Y i Nµ Y, σ ). Then sampling distributions of n 1/ X µ X )/s p and n 1/ Ȳ µ Y )/s p are identically distributed as N, σ /s p) with s p/σ Gν/, /ν), where ν = n and { n } νs p = 1 ρ ) 1 X i X) n ρ X i X)Y n i Ȳ ) + Y i Ȳ ). i=1 i=1 i=1 s S )
12 494 Hea-Jung Kim Above result and the definition of the PS a,b) θ, Σ, g ) ν ) distributionalso see Example 6) immediately gives n 1 n 1 X µ X s p Ȳ > µy ST, 1, ν, λ), X µ X s p Ȳ < µy ST, 1, ν, λ), where λ = ρ 1 ρ. From 4.14), we obtain the unconditional pdf of Z X = n 1/ X µ X )/s p that is ) ) λz X ν + 1 λz hz X ) = f ν z X )F ν+1 X ν f ν + z ν z X )F ν+1 X ν + z X = f ν z X ), where f ν z X ) is the pdf of a standard t ν distribution. Thus, we see that irrespective of the value of ρ ρ 1), Z X t n. Similar argument gives Z Y = n 1/ Ȳ µ Y )/s p t n. These imply that, Z X and Z Y are pivotal quantities for µ X and µ Y for the bivariate normal population with common variance σ and known ρ A Kalman filtering A well-known property of the normal distribution is that, if Y is NZ, σ ) where a priori Z is a normal variable, then the a posteriori distribution Z is still normal. Kim 7) showed that analogous fact is true if a priori Z has probability density function in 3.5). Under this prior, some simple algebra shows that the a posteriori density function of Z given Y = y is still of type 3.5) with θ 1, σ 1, λ, ua), ub)) replaced by y/τ + θ 1 /σ1 1/τ + 1/σ1, 1 σ ) 1 ) 1, λ τ 1 + σ 1 τ, { λ 1 ua) + λσ } 1y θ 1 ) {1 σ1 + τ + λ τ σ1 + τ { λ 1 ub) + λσ } 1y θ 1 ) {1 σ1 + τ + λ τ σ1 + τ } 1, } 1. Note that the parameter λ shrinks towards, independently of y and that the updating formulae of the first two parameters are the same as those of the normal case. Consider now the Kalman filtering setting Z t = ρz t 1 + ε t, Y t = Z t + η t, t = 1,,..., where {ε t } is white noise N, σ ε) and {η t } is white noise N, σ η), with {ε t } independent of {η t }. If the initial prior of Z is normal, then all subsequent posterior distributions of
13 Perturbed Distribution Associated with Truncated Elliptical Model 495 Z t given Y 1,..., Y t ) are still normal. An analogous property holds for the distribution 3.5): if the initial prior distributions of Z is of type 3.5), then all subsequent posterior distributions of Z t given Y 1,..., Y t ) are again of type 3.5). This fact follows from the above conjugacy property and the application of Theorem 4.1. When Z is a random variable with density function 3.) and ε is an independent N, σ ε) variable, Z + ε has density function of type 3.5) by Theorem Conclusion This paper has presented a new class of perturbed elliptical distributions. To form the class of distributions, we considered a conditioning method to the bivariate elliptical distributions. This introduces yet other perturb function, inducing perturbation of the symmetry with univariate elliptical distribution, that brings additional flexibility of modeling skewed and heavy tailed distribution. The results obtained in this paper extend many properties of the bivariate elliptical distributions in a nontrivial way. Several properties of the class are studied, especially for the perturbed normal and perturbed scale mixture of normal distributions. The study shows that we have at hand a class of distributions with following properties: i) strict inclusion of the normal, skew-normal and their scale mixture densities, ii) mathematical tractability and iii) wide applicability in solving statistical problems as demonstrated by Section 5. References Arellano-Valle, R. B., Branco, M. D. and Genton, M. G. 6). A unified view on skewed distributions arising from selections, The Canadian Journal of Statistics/La revue Canadienne de Ststistique, 34, Arnold, B. C., Beaver, R. J., Groeneveld, R. A. and Meeker, W. Q. 1993). The nontruncated marginal of a truncated bivariate normal distribution, Psychometrika, 58, Azzalini, A. 1985). A class of distributions which includes the normal ones, Scandinavian Journal of Statistics, 1, Azzalini, A. and Capitanio, A. 3). Distributions generated by perturbation of symmetry with emphasis on a multivariate skew t distribution, Journal of the Royal Statistical Society, Ser. B, 35, Branco, M. D. and Dey, D. K. 1). A general class of multivariate skew-elliptical distributions, Journal of Multivariate Analysis, 79, Chen, M. H. and Dey, D. K. 1998). Bayesian modeling of correlated binary responses via scale mixture of multivariate normal link functions, Sankhyã, 6, Devroye, L. 1986). Non-Uniform Random Variate Generation, Springer Verlag, New York. Fang, K. T., Kotz, S. and Ng, K. W. 199). Symmetric Multivariate and Related Distributions, Chapman & Hall/CRC, New York. Fang, K. T. and Zhang, Y. T. 199). Generalized Multivariate Analysis, Springer- Verlag, New York.
14 496 Hea-Jung Kim Henze, N. 1986). A probabilistic representation of the Skew-normal distribution, Scandinavian Journal of Statistics, 13, Johnson, N. L. and Kotz, S. 197). Distributions in Statistics: Continuous Multivariate Distributions, John Wiley & Sons, New York. Kim, H. J. ). Binary regression with a class of skewed t link models, Communications in Statistics-Theory and Methods, 31, Kim, H. J. 7). A class of weighted normal distributions and its variants useful for inequality constrained analysis, Statistics, 41, Kim, H. J. 8). A class of weighted multivariate normal distributions and its properties, Journal of the Multivariate Analysis, In press, doi:1.116/j.jmva Ma, Y. and Genton, M. G. 4). A flexible class of skew-symmetric distributions, Scandinavian Journal of Statistics, 31, Rudy, D. A. ). Intermediate Microeconomic Theory, Digital Authoring Resources, Denver. [Received March 8, Accepted April 8]
Tail dependence in bivariate skew-normal and skew-t distributions
Tail dependence in bivariate skew-normal and skew-t distributions Paola Bortot Department of Statistical Sciences - University of Bologna paola.bortot@unibo.it Abstract: Quantifying dependence between
More informationA NEW CLASS OF SKEW-NORMAL DISTRIBUTIONS
A NEW CLASS OF SKEW-NORMAL DISTRIBUTIONS Reinaldo B. Arellano-Valle Héctor W. Gómez Fernando A. Quintana November, 2003 Abstract We introduce a new family of asymmetric normal distributions that contains
More informationarxiv: v1 [math.pr] 10 Oct 2017
Yet another skew-elliptical family but of a different kind: return to Lemma 1 arxiv:1710.03494v1 [math.p] 10 Oct 017 Adelchi Azzalini Dipartimento di Scienze Statistiche Università di Padova Italia Giuliana
More informationIdentifiability problems in some non-gaussian spatial random fields
Chilean Journal of Statistics Vol. 3, No. 2, September 2012, 171 179 Probabilistic and Inferential Aspects of Skew-Symmetric Models Special Issue: IV International Workshop in honour of Adelchi Azzalini
More informationMultivariate Distributions
IEOR E4602: Quantitative Risk Management Spring 2016 c 2016 by Martin Haugh Multivariate Distributions We will study multivariate distributions in these notes, focusing 1 in particular on multivariate
More informationA NOTE ON GENERALIZED ALPHA-SKEW-NORMAL DISTRIBUTION. A.H. Handam Department of Mathematics Al Al-Bayt University P.O. Box , Al Mafraq, JORDAN
International Journal of Pure and Applied Mathematics Volume 74 No. 4 2012, 491-496 ISSN: 1311-8080 printed version url: http://www.ijpam.eu PA ijpam.eu A NOTE ON GENERALIZED ALPHA-SKEW-NORMAL DISTRIBUTION
More informationSome properties of skew-symmetric distributions
Noname manuscript No. (will be inserted by the editor) Some properties of skew-symmetric distributions Adelchi Azzalini Giuliana Regoli 21st December 2010 Revised May 2011 Abstract The family of skew-symmetric
More informationModulation of symmetric densities
1 Modulation of symmetric densities 1.1 Motivation This book deals with a formulation for the construction of continuous probability distributions and connected statistical aspects. Before we begin, a
More informationChapter 5 continued. Chapter 5 sections
Chapter 5 sections Discrete univariate distributions: 5.2 Bernoulli and Binomial distributions Just skim 5.3 Hypergeometric distributions 5.4 Poisson distributions Just skim 5.5 Negative Binomial distributions
More informationBias-corrected Estimators of Scalar Skew Normal
Bias-corrected Estimators of Scalar Skew Normal Guoyi Zhang and Rong Liu Abstract One problem of skew normal model is the difficulty in estimating the shape parameter, for which the maximum likelihood
More informationJohn Geweke Department of Economics University of Minnesota Minneapolis, MN First draft: April, 1991
Efficient Simulation from the Multivariate Normal and Student-t Distributions Subject to Linear Constraints and the Evaluation of Constraint Probabilities John Geweke Department of Economics University
More informationDependence. Practitioner Course: Portfolio Optimization. John Dodson. September 10, Dependence. John Dodson. Outline.
Practitioner Course: Portfolio Optimization September 10, 2008 Before we define dependence, it is useful to define Random variables X and Y are independent iff For all x, y. In particular, F (X,Y ) (x,
More informationA Skewed Look at Bivariate and Multivariate Order Statistics
A Skewed Look at Bivariate and Multivariate Order Statistics Prof. N. Balakrishnan Dept. of Mathematics & Statistics McMaster University, Canada bala@mcmaster.ca p. 1/4 Presented with great pleasure as
More informationThe Mixture Approach for Simulating New Families of Bivariate Distributions with Specified Correlations
The Mixture Approach for Simulating New Families of Bivariate Distributions with Specified Correlations John R. Michael, Significance, Inc. and William R. Schucany, Southern Methodist University The mixture
More informationMotivation Scale Mixutres of Normals Finite Gaussian Mixtures Skew-Normal Models. Mixture Models. Econ 690. Purdue University
Econ 690 Purdue University In virtually all of the previous lectures, our models have made use of normality assumptions. From a computational point of view, the reason for this assumption is clear: combined
More informationYUN WU. B.S., Gui Zhou University of Finance and Economics, 1996 A REPORT. Submitted in partial fulfillment of the requirements for the degree
A SIMULATION STUDY OF THE ROBUSTNESS OF HOTELLING S T TEST FOR THE MEAN OF A MULTIVARIATE DISTRIBUTION WHEN SAMPLING FROM A MULTIVARIATE SKEW-NORMAL DISTRIBUTION by YUN WU B.S., Gui Zhou University of
More informationOn Reparametrization and the Gibbs Sampler
On Reparametrization and the Gibbs Sampler Jorge Carlos Román Department of Mathematics Vanderbilt University James P. Hobert Department of Statistics University of Florida March 2014 Brett Presnell Department
More informationEstimation of state space models with skewed shocks
Estimation of state space models with skewed shocks Grzegorz Koloch April 30, 2012 Abstract This paper provides analytical formulae for the likelihood function, predictive densities and the filtration
More informationEstimation of parametric functions in Downton s bivariate exponential distribution
Estimation of parametric functions in Downton s bivariate exponential distribution George Iliopoulos Department of Mathematics University of the Aegean 83200 Karlovasi, Samos, Greece e-mail: geh@aegean.gr
More informationNotes on a skew-symmetric inverse double Weibull distribution
Journal of the Korean Data & Information Science Society 2009, 20(2), 459 465 한국데이터정보과학회지 Notes on a skew-symmetric inverse double Weibull distribution Jungsoo Woo 1 Department of Statistics, Yeungnam
More informationNonparametric Identification of a Binary Random Factor in Cross Section Data - Supplemental Appendix
Nonparametric Identification of a Binary Random Factor in Cross Section Data - Supplemental Appendix Yingying Dong and Arthur Lewbel California State University Fullerton and Boston College July 2010 Abstract
More informationIntroduction to Normal Distribution
Introduction to Normal Distribution Nathaniel E. Helwig Assistant Professor of Psychology and Statistics University of Minnesota (Twin Cities) Updated 17-Jan-2017 Nathaniel E. Helwig (U of Minnesota) Introduction
More informationLatent Variable Models for Binary Data. Suppose that for a given vector of explanatory variables x, the latent
Latent Variable Models for Binary Data Suppose that for a given vector of explanatory variables x, the latent variable, U, has a continuous cumulative distribution function F (u; x) and that the binary
More informationOn the Conditional Distribution of the Multivariate t Distribution
On the Conditional Distribution of the Multivariate t Distribution arxiv:604.0056v [math.st] 2 Apr 206 Peng Ding Abstract As alternatives to the normal distributions, t distributions are widely applied
More informationDynamic System Identification using HDMR-Bayesian Technique
Dynamic System Identification using HDMR-Bayesian Technique *Shereena O A 1) and Dr. B N Rao 2) 1), 2) Department of Civil Engineering, IIT Madras, Chennai 600036, Tamil Nadu, India 1) ce14d020@smail.iitm.ac.in
More informationThe Use of Copulas to Model Conditional Expectation for Multivariate Data
The Use of Copulas to Model Conditional Expectation for Multivariate Data Kääri, Meelis; Selart, Anne; Kääri, Ene University of Tartu, Institute of Mathematical Statistics J. Liivi 2 50409 Tartu, Estonia
More informationJMASM25: Computing Percentiles of Skew- Normal Distributions
Journal of Modern Applied Statistical Methods Volume 5 Issue Article 3 --005 JMASM5: Computing Percentiles of Skew- Normal Distributions Sikha Bagui The University of West Florida, Pensacola, bagui@uwf.edu
More informationQuadratic forms in skew normal variates
J. Math. Anal. Appl. 73 (00) 558 564 www.academicpress.com Quadratic forms in skew normal variates Arjun K. Gupta a,,1 and Wen-Jang Huang b a Department of Mathematics and Statistics, Bowling Green State
More informationCharacterizations of Weibull Geometric Distribution
Journal of Statistical Theory and Applications Volume 10, Number 4, 2011, pp. 581-590 ISSN 1538-7887 Characterizations of Weibull Geometric Distribution G. G. Hamedani Department of Mathematics, Statistics
More informationThe skew-normal distribution and related multivariate families. Adelchi Azzalini. Università di Padova. NordStat 2004: Jyväskylä, 6 10 June 2004
The skew-normal distribution and related multivariate families Adelchi Azzalini Università di Padova NordStat 2004: Jyväskylä, 6 10 June 2004 A. Azzalini, NordStat 2004 p.1/53 Overview Background & motivation
More informationAn Introduction to Bayesian Linear Regression
An Introduction to Bayesian Linear Regression APPM 5720: Bayesian Computation Fall 2018 A SIMPLE LINEAR MODEL Suppose that we observe explanatory variables x 1, x 2,..., x n and dependent variables y 1,
More informationSample mean, covariance and T 2 statistic of the skew elliptical model
Journal of Multivariate Analysis 97 (2006) 1675 1690 www.elsevier.com/locate/jmva Sample mean, covariance and T 2 statistic of the skew elliptical model B.Q. Fang,1 Institute of Applied Mathematics, Academy
More informationVariational Principal Components
Variational Principal Components Christopher M. Bishop Microsoft Research 7 J. J. Thomson Avenue, Cambridge, CB3 0FB, U.K. cmbishop@microsoft.com http://research.microsoft.com/ cmbishop In Proceedings
More informationBayesian linear regression
Bayesian linear regression Linear regression is the basis of most statistical modeling. The model is Y i = X T i β + ε i, where Y i is the continuous response X i = (X i1,..., X ip ) T is the corresponding
More informationFormulas for probability theory and linear models SF2941
Formulas for probability theory and linear models SF2941 These pages + Appendix 2 of Gut) are permitted as assistance at the exam. 11 maj 2008 Selected formulae of probability Bivariate probability Transforms
More informationChapter 5. Chapter 5 sections
1 / 43 sections Discrete univariate distributions: 5.2 Bernoulli and Binomial distributions Just skim 5.3 Hypergeometric distributions 5.4 Poisson distributions Just skim 5.5 Negative Binomial distributions
More informationOnline tests of Kalman filter consistency
Tampere University of Technology Online tests of Kalman filter consistency Citation Piché, R. (216). Online tests of Kalman filter consistency. International Journal of Adaptive Control and Signal Processing,
More informationSKEW-NORMALITY IN STOCHASTIC FRONTIER ANALYSIS
SKEW-NORMALITY IN STOCHASTIC FRONTIER ANALYSIS J. Armando Domínguez-Molina, Graciela González-Farías and Rogelio Ramos-Quiroga Comunicación Técnica No I-03-18/06-10-003 PE/CIMAT 1 Skew-Normality in Stochastic
More informationOn Bivariate Transformation of Scale Distributions
On Bivariate Transformation of Scale Distributions M.C. Jones The Open University, U.K. ABSTRACT Elsewhere, I have promoted the notion of (univariate continuous) transformation of scale (ToS) distributions
More informationCONDITIONALLY SPECIFIED MULTIVARIATE SKEWED DISTRIBUTIONS
Sankhyā : The Indian Journal of Statistics San Antonio Conference: selected articles, Volume 64, Series A, Pt., pp 6-6 CONDITIONALLY SPECIFIED MULTIVARIATE SKEWED DISTRIBUTIONS By BARRY C. ARNOLD, University
More informationLecture 21: Convergence of transformations and generating a random variable
Lecture 21: Convergence of transformations and generating a random variable If Z n converges to Z in some sense, we often need to check whether h(z n ) converges to h(z ) in the same sense. Continuous
More informationIntroduction of Shape/Skewness Parameter(s) in a Probability Distribution
Journal of Probability and Statistical Science 7(2), 153-171, Aug. 2009 Introduction of Shape/Skewness Parameter(s) in a Probability Distribution Rameshwar D. Gupta University of New Brunswick Debasis
More informationBayesian inference for multivariate skew-normal and skew-t distributions
Bayesian inference for multivariate skew-normal and skew-t distributions Brunero Liseo Sapienza Università di Roma Banff, May 2013 Outline Joint research with Antonio Parisi (Roma Tor Vergata) 1. Inferential
More informationShape mixtures of multivariate skew-normal distributions
Journal of Multivariate Analysis 100 2009 91 101 Contents lists available at ScienceDirect Journal of Multivariate Analysis journal homepage: www.elsevier.com/locate/jmva Shape mixtures of multivariate
More informationStatistical Methods in Particle Physics
Statistical Methods in Particle Physics. Probability Distributions Prof. Dr. Klaus Reygers (lectures) Dr. Sebastian Neubert (tutorials) Heidelberg University WS 07/8 Gaussian g(x; µ, )= p exp (x µ) https://en.wikipedia.org/wiki/normal_distribution
More informationDistributions generated by perturbation of symmetry with emphasis on a multivariate skew distribution
Distributions generated by perturbation of symmetry with emphasis on a multivariate skew distribution ADELCHI AZZALINI Dipartimento di Scienze Statistiche Università di Padova azzalini@stat.unipd.it ANTONELLA
More informationA new construction for skew multivariate distributions
Journal of Multivariate Analysis 95 005 33 344 www.elsevier.com/locate/jmva A new construction for skew multivariate distributions DipakK. Dey, Junfeng Liu Department of Statistics, University of Connecticut,
More informationEmpirical Bayes Estimation in Multiple Linear Regression with Multivariate Skew-Normal Distribution as Prior
Journal of Mathematical Extension Vol. 5, No. 2 (2, (2011, 37-50 Empirical Bayes Estimation in Multiple Linear Regression with Multivariate Skew-Normal Distribution as Prior M. Khounsiavash Islamic Azad
More informationStat 5101 Notes: Brand Name Distributions
Stat 5101 Notes: Brand Name Distributions Charles J. Geyer September 5, 2012 Contents 1 Discrete Uniform Distribution 2 2 General Discrete Uniform Distribution 2 3 Uniform Distribution 3 4 General Uniform
More informationLecture 25: Review. Statistics 104. April 23, Colin Rundel
Lecture 25: Review Statistics 104 Colin Rundel April 23, 2012 Joint CDF F (x, y) = P [X x, Y y] = P [(X, Y ) lies south-west of the point (x, y)] Y (x,y) X Statistics 104 (Colin Rundel) Lecture 25 April
More informationGeometric Skew-Normal Distribution
Debasis Kundu Arun Kumar Chair Professor Department of Mathematics & Statistics Indian Institute of Technology Kanpur Part of this work is going to appear in Sankhya, Ser. B. April 11, 2014 Outline 1 Motivation
More informationSTATISTICS SYLLABUS UNIT I
STATISTICS SYLLABUS UNIT I (Probability Theory) Definition Classical and axiomatic approaches.laws of total and compound probability, conditional probability, Bayes Theorem. Random variable and its distribution
More informationGaussian processes. Chuong B. Do (updated by Honglak Lee) November 22, 2008
Gaussian processes Chuong B Do (updated by Honglak Lee) November 22, 2008 Many of the classical machine learning algorithms that we talked about during the first half of this course fit the following pattern:
More informationIntroduction to Probabilistic Graphical Models: Exercises
Introduction to Probabilistic Graphical Models: Exercises Cédric Archambeau Xerox Research Centre Europe cedric.archambeau@xrce.xerox.com Pascal Bootcamp Marseille, France, July 2010 Exercise 1: basics
More informationMultivariate Normal-Laplace Distribution and Processes
CHAPTER 4 Multivariate Normal-Laplace Distribution and Processes The normal-laplace distribution, which results from the convolution of independent normal and Laplace random variables is introduced by
More informationLecture 2: Repetition of probability theory and statistics
Algorithms for Uncertainty Quantification SS8, IN2345 Tobias Neckel Scientific Computing in Computer Science TUM Lecture 2: Repetition of probability theory and statistics Concept of Building Block: Prerequisites:
More informationConsistent Bivariate Distribution
A Characterization of the Normal Conditional Distributions MATSUNO 79 Therefore, the function ( ) = G( : a/(1 b2)) = N(0, a/(1 b2)) is a solu- tion for the integral equation (10). The constant times of
More informationINFERENCE FOR MULTIPLE LINEAR REGRESSION MODEL WITH EXTENDED SKEW NORMAL ERRORS
Pak. J. Statist. 2016 Vol. 32(2), 81-96 INFERENCE FOR MULTIPLE LINEAR REGRESSION MODEL WITH EXTENDED SKEW NORMAL ERRORS A.A. Alhamide 1, K. Ibrahim 1 M.T. Alodat 2 1 Statistics Program, School of Mathematical
More informationNew Family of the t Distributions for Modeling Semicircular Data
Communications of the Korean Statistical Society Vol. 15, No. 5, 008, pp. 667 674 New Family of the t Distributions for Modeling Semicircular Data Hyoung-Moon Kim 1) Abstract We develop new family of the
More informationAsymptotic Statistics-III. Changliang Zou
Asymptotic Statistics-III Changliang Zou The multivariate central limit theorem Theorem (Multivariate CLT for iid case) Let X i be iid random p-vectors with mean µ and and covariance matrix Σ. Then n (
More informationA study of generalized skew-normal distribution
Statistics, Vol. 00, No. 0, Month 2012, 1 12 0.8CE: VLD/SP QA: Coll: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47
More information4. CONTINUOUS RANDOM VARIABLES
IA Probability Lent Term 4 CONTINUOUS RANDOM VARIABLES 4 Introduction Up to now we have restricted consideration to sample spaces Ω which are finite, or countable; we will now relax that assumption We
More informationStudent-t Process as Alternative to Gaussian Processes Discussion
Student-t Process as Alternative to Gaussian Processes Discussion A. Shah, A. G. Wilson and Z. Gharamani Discussion by: R. Henao Duke University June 20, 2014 Contributions The paper is concerned about
More informationMarkov Chain Monte Carlo (MCMC)
Markov Chain Monte Carlo (MCMC Dependent Sampling Suppose we wish to sample from a density π, and we can evaluate π as a function but have no means to directly generate a sample. Rejection sampling can
More informationmatrix-free Elements of Probability Theory 1 Random Variables and Distributions Contents Elements of Probability Theory 2
Short Guides to Microeconometrics Fall 2018 Kurt Schmidheiny Unversität Basel Elements of Probability Theory 2 1 Random Variables and Distributions Contents Elements of Probability Theory matrix-free 1
More informationSome Approximations of the Logistic Distribution with Application to the Covariance Matrix of Logistic Regression
Working Paper 2013:9 Department of Statistics Some Approximations of the Logistic Distribution with Application to the Covariance Matrix of Logistic Regression Ronnie Pingel Working Paper 2013:9 June
More informationSkewed Reflected Distributions Generated by the Laplace Kernel
AUSTRIAN JOURNAL OF STATISTICS Volume 38 (009), Number 1, 45 58 Skewed Reflected Distributions Generated by the Laplace Kernel M. Masoom Ali 1, Manisha Pal and Jungsoo Woo 3 1 Dept. of Mathematical Sciences,
More informationSELECTING THE NORMAL POPULATION WITH THE SMALLEST COEFFICIENT OF VARIATION
SELECTING THE NORMAL POPULATION WITH THE SMALLEST COEFFICIENT OF VARIATION Ajit C. Tamhane Department of IE/MS and Department of Statistics Northwestern University, Evanston, IL 60208 Anthony J. Hayter
More informationChris Bishop s PRML Ch. 8: Graphical Models
Chris Bishop s PRML Ch. 8: Graphical Models January 24, 2008 Introduction Visualize the structure of a probabilistic model Design and motivate new models Insights into the model s properties, in particular
More informationSTA 2201/442 Assignment 2
STA 2201/442 Assignment 2 1. This is about how to simulate from a continuous univariate distribution. Let the random variable X have a continuous distribution with density f X (x) and cumulative distribution
More informationFall 2017 STAT 532 Homework Peter Hoff. 1. Let P be a probability measure on a collection of sets A.
1. Let P be a probability measure on a collection of sets A. (a) For each n N, let H n be a set in A such that H n H n+1. Show that P (H n ) monotonically converges to P ( k=1 H k) as n. (b) For each n
More informationPart IA Probability. Definitions. Based on lectures by R. Weber Notes taken by Dexter Chua. Lent 2015
Part IA Probability Definitions Based on lectures by R. Weber Notes taken by Dexter Chua Lent 2015 These notes are not endorsed by the lecturers, and I have modified them (often significantly) after lectures.
More informationOutline Lecture 2 2(32)
Outline Lecture (3), Lecture Linear Regression and Classification it is our firm belief that an understanding of linear models is essential for understanding nonlinear ones Thomas Schön Division of Automatic
More informationDependence. MFM Practitioner Module: Risk & Asset Allocation. John Dodson. September 11, Dependence. John Dodson. Outline.
MFM Practitioner Module: Risk & Asset Allocation September 11, 2013 Before we define dependence, it is useful to define Random variables X and Y are independent iff For all x, y. In particular, F (X,Y
More informationGaussian kernel GARCH models
Gaussian kernel GARCH models Xibin (Bill) Zhang and Maxwell L. King Department of Econometrics and Business Statistics Faculty of Business and Economics 7 June 2013 Motivation A regression model is often
More informationA New Class of Positively Quadrant Dependent Bivariate Distributions with Pareto
International Mathematical Forum, 2, 27, no. 26, 1259-1273 A New Class of Positively Quadrant Dependent Bivariate Distributions with Pareto A. S. Al-Ruzaiza and Awad El-Gohary 1 Department of Statistics
More informationBayesian data analysis in practice: Three simple examples
Bayesian data analysis in practice: Three simple examples Martin P. Tingley Introduction These notes cover three examples I presented at Climatea on 5 October 0. Matlab code is available by request to
More information1 Introduction. P (n = 1 red ball drawn) =
Introduction Exercises and outline solutions. Y has a pack of 4 cards (Ace and Queen of clubs, Ace and Queen of Hearts) from which he deals a random of selection 2 to player X. What is the probability
More informationElements of Probability Theory
Short Guides to Microeconometrics Fall 2016 Kurt Schmidheiny Unversität Basel Elements of Probability Theory Contents 1 Random Variables and Distributions 2 1.1 Univariate Random Variables and Distributions......
More informationFriedrich-Alexander-Universität. Erlangen-Nürnberg
Friedrich-Alexander-Universität Erlangen-Nürnberg Wirtschafts-und Sozialwissenschaftliche Fakultät Diskussionspapier 63 / 2004 The L Distribution and Skew Generalizations Matthias Fischer Lehrstuhl für
More informationRandom vectors X 1 X 2. Recall that a random vector X = is made up of, say, k. X k. random variables.
Random vectors Recall that a random vector X = X X 2 is made up of, say, k random variables X k A random vector has a joint distribution, eg a density f(x), that gives probabilities P(X A) = f(x)dx Just
More informationUniform Correlation Mixture of Bivariate Normal Distributions and. Hypercubically-contoured Densities That Are Marginally Normal
Uniform Correlation Mixture of Bivariate Normal Distributions and Hypercubically-contoured Densities That Are Marginally Normal Kai Zhang Department of Statistics and Operations Research University of
More informationOn Bayesian Inference with Conjugate Priors for Scale Mixtures of Normal Distributions
Journal of Applied Probability & Statistics Vol. 5, No. 1, xxx xxx c 2010 Dixie W Publishing Corporation, U. S. A. On Bayesian Inference with Conjugate Priors for Scale Mixtures of Normal Distributions
More informationProbability Theory and Statistics. Peter Jochumzen
Probability Theory and Statistics Peter Jochumzen April 18, 2016 Contents 1 Probability Theory And Statistics 3 1.1 Experiment, Outcome and Event................................ 3 1.2 Probability............................................
More informationMultiple Random Variables
Multiple Random Variables This Version: July 30, 2015 Multiple Random Variables 2 Now we consider models with more than one r.v. These are called multivariate models For instance: height and weight An
More informationA NOTE ON A DISTRIBUTION OF WEIGHTED SUMS OF I.I.D. RAYLEIGH RANDOM VARIABLES
Sankhyā : The Indian Journal of Statistics 1998, Volume 6, Series A, Pt. 2, pp. 171-175 A NOTE ON A DISTRIBUTION OF WEIGHTED SUMS OF I.I.D. RAYLEIGH RANDOM VARIABLES By P. HITCZENKO North Carolina State
More informationChapter 3 sections. SKIP: 3.10 Markov Chains. SKIP: pages Chapter 3 - continued
Chapter 3 sections Chapter 3 - continued 3.1 Random Variables and Discrete Distributions 3.2 Continuous Distributions 3.3 The Cumulative Distribution Function 3.4 Bivariate Distributions 3.5 Marginal Distributions
More informationMultivariate Distribution Models
Multivariate Distribution Models Model Description While the probability distribution for an individual random variable is called marginal, the probability distribution for multiple random variables is
More informationTail negative dependence and its applications for aggregate loss modeling
Tail negative dependence and its applications for aggregate loss modeling Lei Hua Division of Statistics Oct 20, 2014, ISU L. Hua (NIU) 1/35 1 Motivation 2 Tail order Elliptical copula Extreme value copula
More informationHow to select a good vine
Universitetet i Oslo ingrihaf@math.uio.no International FocuStat Workshop on Focused Information Criteria and Related Themes, May 9-11, 2016 Copulae Regular vines Model selection and reduction Limitations
More informationPhenomena in high dimensions in geometric analysis, random matrices, and computational geometry Roscoff, France, June 25-29, 2012
Phenomena in high dimensions in geometric analysis, random matrices, and computational geometry Roscoff, France, June 25-29, 202 BOUNDS AND ASYMPTOTICS FOR FISHER INFORMATION IN THE CENTRAL LIMIT THEOREM
More informationChapter 3 sections. SKIP: 3.10 Markov Chains. SKIP: pages Chapter 3 - continued
Chapter 3 sections 3.1 Random Variables and Discrete Distributions 3.2 Continuous Distributions 3.3 The Cumulative Distribution Function 3.4 Bivariate Distributions 3.5 Marginal Distributions 3.6 Conditional
More information01 Probability Theory and Statistics Review
NAVARCH/EECS 568, ROB 530 - Winter 2018 01 Probability Theory and Statistics Review Maani Ghaffari January 08, 2018 Last Time: Bayes Filters Given: Stream of observations z 1:t and action data u 1:t Sensor/measurement
More informationMA/ST 810 Mathematical-Statistical Modeling and Analysis of Complex Systems
MA/ST 810 Mathematical-Statistical Modeling and Analysis of Complex Systems Review of Basic Probability The fundamentals, random variables, probability distributions Probability mass/density functions
More informationProbability. Table of contents
Probability Table of contents 1. Important definitions 2. Distributions 3. Discrete distributions 4. Continuous distributions 5. The Normal distribution 6. Multivariate random variables 7. Other continuous
More informationPh.D. Qualifying Exam Friday Saturday, January 6 7, 2017
Ph.D. Qualifying Exam Friday Saturday, January 6 7, 2017 Put your solution to each problem on a separate sheet of paper. Problem 1. (5106) Let X 1, X 2,, X n be a sequence of i.i.d. observations from a
More informationPattern Recognition and Machine Learning. Bishop Chapter 2: Probability Distributions
Pattern Recognition and Machine Learning Chapter 2: Probability Distributions Cécile Amblard Alex Kläser Jakob Verbeek October 11, 27 Probability Distributions: General Density Estimation: given a finite
More informationSpring 2012 Math 541B Exam 1
Spring 2012 Math 541B Exam 1 1. A sample of size n is drawn without replacement from an urn containing N balls, m of which are red and N m are black; the balls are otherwise indistinguishable. Let X denote
More informationOn 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 informationTail dependence for two skew slash distributions
1 Tail dependence for two skew slash distributions Chengxiu Ling 1, Zuoxiang Peng 1 Faculty of Business and Economics, University of Lausanne, Extranef, UNIL-Dorigny, 115 Lausanne, Switzerland School of
More information