TESTING FOR COINTEGRATION RANK USING BAYES FACTORS. Katsuhiro Sugita WARWICK ECONOMIC RESEARCH PAPERS
|
|
- Marjorie Rice
- 5 years ago
- Views:
Transcription
1 TESTING FOR COINTEGRATION RANK USING BAYES FACTORS Katsuhiro Sugita No 5 WARWICK ECONOMIC RESEARCH PAPERS DEPARTMENT OF ECONOMICS
2 Testing for Cointegration Rank Using Bayes Factors Katsuhiro Sugita Λ July 00 Department of Economics University of Warwick, UK Abstract This paper proposes Bayesian methods for estimating the cointegration rank using Bayes factors We consider natural conjugate priors for computing Bayes factors First, we estimate the cointegrating vectors for each possible rank Then, we compute the Bayes factors for each rank against 0 rank Monte Carlo simulations show that using Bayes factor with conjugate priors produces fairly good results We apply the method to demand for money in the US Key words: Cointegration; MCMC; Bayes factor JEL classification: C11; C1; C 1 Introduction This paper introduces Bayesian analysis of cointegrated VAR systems Several researchers have proposed Bayesian approach to cointegrated VAR systems These include Kleibergen and van Dijk (199), who proposed using a Jeffrey's prior instead of diffuse prior for the cointegrating vectors since the marginal posteriors may be nonintegrable with reduced rank of cointegrating vector Geweke(199) developed general methods for Bayesian inference with noninformative reference priors in the reduced rank regression model Kleibergen and Paap (1999) use a singular value decomposition of the unrestricted long-run multiplier matrix, Π, for identification of the cointegrating vectors and for Bayesian posterior odds analysis of the rank of Π Bauwens and Lubrano (199) reduce the ECM to the form of a multivariate regression model to identify the parameters The cointegrating rank is assumed to be known a priori, based on a theoretical economic model that defines equilibrium economic relations Λ The author thanks two anonymous referees, Luc Bauwens, Mike Clements, Richard Paap and audiences at the workshop Recent Advances in Bayesian Econometrics (001) in Marseilles for their useful comments on an earlier version of this paper 1
3 If we are interested in identifying the cointegration rank, they suggest checking the plot of the posterior density of the eigenvalues of generated sample Π 0 Π, which are equal to the square of the singular values of Π However, this informal visual inspection gives ambiguous results 1 Bauwens, et al (1999) suggest using the trace test of Johansen, since on the Bayesian side, the topic of selecting the cointegrating rank has not yet given very useful and convincing results (p8) In this paper we propose a simple method of determining of the cointegration rank by Bayes factors The method is very straightforward We consider using the conjugate priors for all parameters If there exist r cointegrating vectors in the system, the adjustment term ff has rank r Applying the Bayes factor to ff, which has r rank, against the null of ff, which has rank 0, for each rank gives the posterior probabilities for its rank The procedure for obtaining the posteriors has some similarities with Bauwens and Lubrano method However, instead of using diffuse priors for all parameters, the conjugate priors are chosen to be able to compute the Bayes factors The plan of this paper is as follows Section presents the prior specifications and derives the posterior densities for estimation of the cointegrated VAR systems In Section the Bayes factors for cointegration rank is introduced Section illustrates Monte Carlo simulations with DGPs of 1000 iterations for each rank to compare the performance of the proposed Bayesian methods with the classical Johansen test In Section 5, an illustrative example of the demand for money in the United Sates is presented Section concludes All computations in this paper are performed using code written by the author with Ox v0 for Linux (Doornik, 1998) Bayesian Inference in Cointegration Analysis In this section we present Bayesian analysis of cointegration Let X t denote an I(1) vector of n-dimensional time series with r linear cointegrating relations, then unrestricted VECM representation with deterministic trend is: X t = μ + ffflz + p 1 X Ψ i X t i + " t (1) h i 0, h i where z = X 0 t 1 t fl = fi 0 ffi P, t = p; p +1;:::;T, p is the number of lags in VAR, and the errors, " t, are assumed N (0; ) and independent over time μ, ", Ψ, ±, ff, andfi are parameters of dimensions n 1; n 1, n n, n n, n r, and n r, respectively Equation (1) can be rewritten in matrix format as: where Y = X +Zfl 0 ff 0 + E = WB + E () 1 Tsurumi and Wago (199) use a highest-posterior-density-region (HPDR) test to Π, then derive the posterior pdfs for singular values to see whether 99% highest-posterior-density-interval (HPDI) contains zero
4 Y = X = X 0 p X 0 p+1 X 0 T 5, Z = X 0 p 1 p X 0 p p +1 X 0 T 1 1 X 0 p 1 X X 0 p X 0 1 X 0 T 1 X 0 T p+1 T 5, E = h 5, W = " 0 p " 0 p+1 " 0 T 5, = X Zfl 0 i, B = μ 0 Ψ 0 1 Ψ 0 p 1 " 5, # ff 0 Let m be the number of columns of Y, so that m = T p +1, then X is m (1 + n(p 1)), ((1 + n(p 1)) n), W (m k), where k =1+n(p 1) + r, and B (k n) Thus, equation () represents the multivariate regression format of (1) This representation is a starting point We then describe the prior and likelihood specifications in order to derive posteriors First, we consider the case of applying the conjugate priors for some parameters strategy to select priors is to choose a conjugate prior only for the parameters that are used in computing Bayes factor For other parameters except for the cointegrating vector, we consider non-informative priors The conjugate prior density for B conditional on covariance ± follows a matrix-variate normal distribution with covariance matrix ± Ω A 1 of the form p(b j ±) /j±j k= jaj n= exp» Our 1 trφ ± 1 (B P ) 0 A (B P ) Ψ () where A is (k k) PDS and P (k n), k = n(p 1) + r +1 (the number of columns in W ) Choosing hyperparameters, P and A, in () should be careful since these have direct effect on the value of Bayes factor Kleibergen and Paap (199) choose The idea of using data in priors is similar to a g-prior proposed by Zellner (198) For the prior density for the covariance ± in (), we can assign an inverted Wishart ρ 1 tr ± S ff 1 p(±) /jsj h= j±j (h+n+1) exp () where h is a degree of freedom, S an n n PDS The prior for fi can be given as a matrix-variate normal ß (fi) /jqj n= jhj r= exp» 1 tr nq 1 fi fi 0 H fi fi o (5) where fi is a prior mean of fi, Q is r r PDS, H is n n PDS Note that r restrictions for identification are imposed on fi, for example, fi 0 = ( I r fi? 0 ), where fi? is (n r) r unrestricted matrix If we assign r restrictions on fi as I r, then only a part of fi, fi?, follows a matrix-variate normal If we assume that B, ± and fi are mutually independent, then the joint prior of the parameters in () is p(b;fi; ±) / p(bj±)p(fi)p(±) and thus can be derived as The restrictions imposed on fi need not to be I r but can be any r restrictions See Bauwens and Lubrano (199, page 1)
5 » p(b; ±;fi) / ß (fi) jaj n= j±j k+h+n+1 exp 1 trφ ΛΨ ± 1 S +(B P ) 0 A(B P ) () To derive the conditional posterior distributions, we need to derive the likelihood functions The likelihood function for B; ±; and fi is given by: n oio L (Y j B; ±;fi) / j±j t= exp n h± 1 tr 1 bs +(B B) b 0 W 0 W (B B) b () where b B =(W 0 W ) 1 W 0 Y, and b S =(Y W b B) 0 (Y W b B) Next we derive the posteriors from the priors and the likelihood function specified above The joint posterior distribution for the conjugate priors for ff is proportional to the joint prior () times the likelihood function (), thus we have p (B; ±;fi j Y ) / p (B; ±;fi) L (Y j B; ±;fi) / ß (fi) jaj n j±j (t+h+k+n+1)= exp / ß (fi) j±j c exp» 1 tr n± 1 hs +(B P ) 0 A(B P )+b S +(B b B) 0 W 0 W (B b B io h» 1 n± tr 1 S + S b +(P B) b 0 [A 1 +(W 0 W ) 1 ] 1 (P B) b ΛΨΛ +(B B? ) 0 A? (B B? )» 1 trφ ΛΨ ± 1 S? +(B B? ) 0 A? (B B? ) (8) = ß (fi) j±j c exp where c = t + k + h + n +1, A? = A + W 0 W, B? =(A + W 0 W ) 1 (AP + W 0 W b B), and S? = S + b S +(P b B) 0 [A 1 +(W 0 W ) 1 ] 1 (P b B) From (8), the conditional posterior of ± is derived as an inverted Wishart distribution, and the conditional posterior of B as a matrix-variate normal density with covariance, ± Ω A 1?, that is, p(± j fi;y ) /js? j t= j±j (t+h+n+1)= exp p(b j ±;fi;y) /ja? j n= j±j k= exp»» 1 tr ± 1 S? (9) 1 trφ ± 1 (B B? ) 0 A? (B B? ) Ψ (10) Thus, by multiplying (9) and (10), and integrating with respect to ±, we obtain the posterior density of B conditional on fi, which is a matrix-variate Student-t form, p(b j fi;y ) /js? j t= ja? j n= js? +(B B? ) 0 A? (B B? )j (t+k)= (11) The joint posterior of B and fi can be derived by integrating (8) with respect to ±,
6 p(b;fi j Y ) / ß (fi) js? +(B B? ) 0 A? (B B? ) j (t+h+k+1)= (1) By integrating (1) with respect to B we obtain the posterior density of the cointegrating vector fi, p(fi j Y ) / ß (fi) j S? j (t+h+1)= j A? j n= (1) The properties of (1) are not known, so that we have to resort to numerical integration techniques as Bauwens and Lubrano (199) use importance sampling to compute poly-t posterior results on parameters Other feasible methods are the Metropolis-Hastings algorithm and the Griddy-Gibbs sampling The Metropolis-Hastings algorithm requires assignment of a good approximating function, the candidate-generating function, to the posterior to draw random numbers, as importance sampling requires the importance function Since the Griddy- Gibbs sampling method does not require such an approximation, we employ the Griddy-Gibbs sampler for estimation of the cointegrating vector as Bauwens and Giot (1998) use the sampler for estimation of two cointegrating vectorsthe Griddy-Gibbs sampler that is proposed by Ritter and Tanner (199) approximates the true cdf of each conditional distribution by a piecewise linear function and then sample from the approximations The disadvantage of this sampling method is that we have to assign proper range and the number of the grid The range should be chosen so that the generated numbers are not truncated Bayes Factors for Cointegration Tests This section introduces the computation of the Bayes factors for cointegration rank Subsection 1 describes briefly the basic concept of the Bayes factors and some computation techniques Subsection presents the computation of the Bayes factors for cointegration rank 1 Bayes Factors and Determination of Rank The Bayes factor, which is defined as the ratio of marginal likelihood of null and alternative hypothesis, has been used for model selection The Bayes factors can be used to construct posterior probabilities for all models that seemed plausible In classical hypothesis test, one model represents the truth and the test is based on a pairwise comparison of the alternative For a detailed discussion of the advantages of Bayesian methods, see Koop and Potter (1999) Kass and Raftery (1995) provide an excellent survey of the Bayes factor Suppose, with data Y and the likelihood functions with the parameters, there are two hypotheses H 0 and H 1 The Bayes factor BF 01 is defined as follows: BF 01 = Pr(Y jh 0) Pr(Y jh 1 ) For more details, consult Chen, et al (000), Evans and Swartz (000) For tutorial for the M-H algorithm, see Chib and Greenberg (1995) 5
7 = R p( jh0 )L(Y j ;H 0 )d R p( jh1 )L(Y j ;H 1 )d (1) With the prior odds, defined as Pr(H 0 )=Pr(H 1 ), we can compute the posterior odds, which are PosteriorOdds 01 = Pr(H 0jY ) Pr(H 1 jy ) = Pr(Y jh 0) Pr(Y jh 1 ) Pr(H 0) Pr(H 1 ) When several models are being considered, the posterior odds yield the posterior probabilities Suppose q models with H 0 ;H 1 ;:::;H q 1 are being considered, and each hypotheses of H 1 ;H ;:::;H q 1 is compared with H 0 Then the posterior probability for model i under H i is (15) Pr(H i jy )= PosteriorOdds i0 P q j=0 PosteriorOdds j0 (1) where PosteriorOdds 00 is defined to be 1 These posterior probabilities are used to select the cointegrating rank, model selection, or as weights for forecasting There are several methods to compute the Bayes factors given in (1) For example, the Laplace approximation method (Tierney and Kadane, 198), or using numerical integration techniques such as importance sampling (Geweke, 1989) or the Metropolis-Hastings algorithm See Kass and Raftery (1995) for details Chib (1995) proposes a simple approach to compute the marginal likelihood from the Gibbs output In the case of nested models computation of the Bayes factor can be simplified by using the generalized Savage-Dickey density ratio, proposed by Verdinelli and Wasserman (1995), which is based on Dickey (191) Suppose we wish to test the null H 0 : ο = ο 0 versus H 1 : ο = ο 0, where ο can be scalar, vector, or matrix With the condition that p(φjο 0 ) = p 0 (Φ), where (Φ;ο) =, then the Bayes factor can be computed with the Savage-Dickey density ratio BF 01 = p(ο 0jY ) p(ο 0 ) The denominator, the marginal prior for ο evaluated at ο = ο 0, in (1) is trivial to calculate The numerator, the marginal posterior for ο evaluated at ο = ο 0, can be calculated by integrating out the other parameters, such as: (1) p(ο 0 jy ) = ' Z 1 N p(ο 0 jφ;y)p(φjy )d NX p(ο 0 jφ (i) ;Y) (18) where Φ (i), i =1;:::;N, are sample draws from the posterior
8 Bayes Factor for Cointegration Rank The Bayes factors are used for model selection, and thus can also be used for rank selection of the cointegration Kleibergen and Paap (1999) propose a cointegration rank test by Bayes factors using a decomposition derived from the singular value decomposition In this subsection, we propose a much simpler method for rank test using Bayes factors In a cointegrated system with n variables which arei(1), if there are r cointegrating vectors, then the error correction term Π has reduced rank of r: Π can be decomposed as products of ff and fi 0, both of which have reduced rank r Since ff is unrestricted and also is given a rank reduction when cointegration exists, we compute the Bayes factor for ff that is against the null ff = 0 for determining the number of the rank using inverted form of (1) for each possible rank (r =0; 1;:::;n) BF rj0 = = R R R R RRR p (ff; fi; ; ±) L (Y j ff; fi; ; ±) dffdfid d± (1=C r ) p (ff; fi; ; ±) j L (Y j ff; fi; ; rank(ff)=0 ±) j rank(ff)=0 dfid d± p(ff 0 =0 r n ) (19) (1=C r ) p(ff 0 =0 r n jy ) where C r = R R R p(ff; fi; ; ±) j rank(ff)=0 reduction of dimension dfid d± is the correction factor that is required for If there exists r cointegrating vectors, the Bayes factor for ff 0 in (19) is the most (r n) unlikely to be zero and thus should have the highest value in Bayes factors for other possible ranks Note that the Bayes factor for rank 0 equals to 1 In case of no cointegration, the Bayes factors for ff 0,wherer = 0, are less than 1 If we assign an equal prior probability toeach (r n) cointegration rank, the posterior probability for each rank can be computed as in (1) Pr(rjY )= BF rj0 P n j=0 BF jj0 (0) where BF 0j0 is defined as 1 The posterior probabilities given by (0) can be used for solutions of the prediction, decision making and inference problems that take account of model uncertainty Generally, the hypothesis that has the highest posterior probability can be selected as the 'best' model, only if it dominates the others Otherwise, analyses will fail to take uncertainty into account To compute the Bayes factors using (19), we use (18) with samples from the posteriors Since ff 0 is a partitioned element of B in () and thus the prior for alpha is a matrix-variate Student-t distribution as shown in (), so the numerator of (19) is: 8 p ff 0 =0 (n r) = ß nr h < jsj ja :1 j n= : where A :1 = A A 1 A 1 11 A 1, A = " ny j=1 h+r+1 j h+1 j 9 = h+r ; jsj (1) # A 11 A 1, A 11 is ((n(p 1) + 1) (n(p 1) + 1)) A 1 A
9 , A 1 ((n(p 1) + 1) r), A 1 (r (n(p 1) + 1)), A (r r) The posterior for alpha, which is a matrix-variate Student-t from (11), can be estimated by (18) as follows: p(ff 0 =0 (n r) jy ) = Z = 1 N p(ff 0 =0jfi;Y )p(fijy )dfi ' 1 N NX ß nr js (i)? j t+h ja (i)?;1 j n js (i)? + B (i)0? A(i)?:1 B(i)? ( ny NX p(ff 0 =0jfi (i) ;Y) t+h+r+1 i t+h+1 i ) j t+r () " # A?11 A?1, A?11 is ((n(p 1) + 1) (n(p 1) + 1)), A?1 A?1 A? ((n(p 1) + " 1) r) #, A?1 (r (n(p 1) + 1)), A? (r r), B? is obtained by partition of B?1 B? as B? =, where B?1 is ((n(p 1) + 1) n), B? (r n) B? where fi (i), S (i)?, and A (i)? are obtained from the i th iteration of the Gibbs sampler, A?:1 = A? A?1 A 1?11 A?1, A? = 1, To compute the value of C r in (19), Chen's method (199) can be used as R RRR p (ff; fi; ; ±) dffdfid d± = C r = Z Z Z p (ff; fi; ; ±) j ff=0 dfid d± = = R R R p (ff; fi; ; ±) jff=0 dfid d± R R R R R R R p (ff; fi; ; ±) dffdfid d± R p (ff; fi; R RRR ; ±) jff=0! (ff j fi; ; ±) dff dfid d± p (ff; fi; ; ±) dffdfid d± = ' R R RR p (ff; fi; ; ±) jff=0! (ff j fi; ; ±) dffdfid d± RRR R p (ff; fi; ; ±) dffdfid d± NX 1 N = 1 N = 1 N = 1 N NX NX NX p ff p (i) ff =0;fi (i) ; (i) j fi (i) ; (i) ; ± (i) (i) ; ± (i) p ff (i) ;fi (i) ; (i) ; ± p ff =0;fi (i) ; (i) ; ± (i) p fi (i) ; (i) ; ± (i) (i) (i) (i) p ff =0; (i) (i) j ± p (i) fi p (i) ± p (i) j ± (i) p fi p ± p ff =0; (i) j ± p (i) j ± (i) () where p (ff =0; j ±) and p ( j ±) in the last line of () are derived from () The Bayes factor for alpha can be obtained by dividing (1) by () and an inverse of () When the posterior probabilities are considered, we assign equal prior probabilities to 8
10 the possible cointegration ranks such that Pr(ff rank=r )=1=(n +1) for r =0; 1;:::;n With these n +1 Bayes factors, we can compute the posterior probabilities for each rank by using (1) Note that the method described above has the disadvantage over other methods such as Johansen's and Kelibergen and Paap's methods Their methods are based on eigenvalues or singular values of the long-run matrix Π, while our method is based on testing for zero restrictions on adjustment parameters The Bayes factors we obtain in our methodology depend upon r restrictions for identification imposed on fi Thus, different choice of the ordering of the variables generates different value of Bayes factors Monte Carlo Simulation To illustrate the performance of Bayesian tests for the rank of cointegration described in the previous section, we perform some Monte Carlo simulations The data generating processes (DGPs) consist of a four-variable VAR with an intercept term having various number of cointegrating vectors (0, 1,, and ) as following: 1 y t = μ + e t y t = μ + y t = μ + y t = μ + 5 y t = μ + where μ = h 0: 0: 0: 0: h 5 0: 0: 0: 0: 0: 0: 0: 0: " 5 0: 0: 0: 0: 0: 0: 0: 0: 0: 0: 0: 0: i y t 1 + e t : 0: 0: 0: 0: 0: 0: 0: 0: 0: 0: 0: 0: 0: 0: 0: 0:1 0:1 0:1 0:1 # y t 1 + e t i 0, and et ο NID(0;I ) 5 y t 1 + e t 5 y t 1 + e t We demonstrate the performance of the test by varying the true rank Each simulation of DGPs for various ranks is repeated 1000 times The sample size t is 150, of which the first 50 are used for the first experiment and the rest of 100 are used for the second experiment We consider a VAR(1) model with a constant term throughout the experiments As noted 9
11 in the previous section, since the testing procedure depends upon the chosen ordering of the variables in the VAR, the ordering of the individual series in y is changed randomly during the simulation experiment The prior parameter specifications for the natural conjugate prior in () are given with P =0and A = (c W 0c W )=T, where c W = X Z b fi and b fi =0(n r) The specification for A is a g-prior proposed by Zellner (198) and used by Kleibergen and Paap In this experiments, we assign =1and 001 to see how this prior specification affects the results We also specify the prior parameters for fi in (5) with fi? = 0, Q = I n, H = fiz 0 Z, where fi = 1=T As for the prior specification for ± in (), we assign S = fix 0 X and h = n +1 For Johansen's LR trace test, the cointegrating rank for each iteration is chosen by p-value with the 5 per cent significance level and then the number of each selected rank is counted over iterations to obtain the rates of selection for each rank Table 1 summarizes the results of Monte Carlo simulation with the sample size t is 50 The values in the columns are the average posterior probabilities of 1,000 iterations for each true rank For each iteration, the Griddy-Gibbs sampling is performed with 5,000 draws and the first 1,000 discarded The column labelled as Pr(r Y) with =1shows the average posterior probabilities when =1 As shown in the table, the highest average posterior probability for each rank indicates the correct rank With the sample size is 50, the overall performances are slightly improved when =0:01, larger prior variance The results of the Johansen tests in this paper are obtained by using Pcfiml class of Ox v0 The source code of the class was modified and recompiled for the simulations 10
12 Table 1: Monte Carlo Results: The Average Posterior Probabilities with T =50 Pr(r Y) Pr(r Y) Johansen's DGP rank r with =1 with =0:01 trace test True rank r = True rank r = True rank r = True rank r = True rank r = The last column shows the results by Johansen's trace test The results show that the test apparently suffers from the shortage of samples especially for higher ranks To improve the finite sample properties for the likelihood ratio test, Johansen (000) proposed using Bartlett correction for a VAR with small sample Increasing the sample size to 100 improves the performances of all tests as shown in Table All highest average posterior probabilities indicate the true rank Larger the prior variance on ff (less value in ), it tends to choose a lower rank 11
13 Table : Monte Carlo Results: The Average Posterior Probabilities with T = 100 DGP rank r Pr(r Y) Pr(r Y) Johansen's with =1 with =0:01 trace test True rank r = True rank r = True rank r = True rank r = True rank r = Illustrative Example: The Demand for Money in the US In this section, we illustrate an example of cointegration analysis using the method that is presented in previous sections The focus is to show the usefulness of our method with a relatively small sample size and to compare two methods of computing Bayes factors and Johansen's likelihood ratio test The example is cointegration test for the demand for money in the United States There are many papers on the estimation of the money demand function, see Goldfeld and Sichel (1990) Atypical money demand equation is m t p t = μ + fl y y t + fl R R t + e t () 1
14 where m t is the log of money, p t is the log of price level, y t is log of income, R t is the nominal interest rate The data used for our analysis is based on the annual data provided by Lucas (1988) The money stock (m t ) is measured by M1 The income (y t ) is the net national product The interest rate (R t ) is the annual rate of the six-month commercial paper We use a part of the range of the original data - from 19 to 1989 With 0 observations, we test cointegration among three variables - (m p), y, and R Before analyzing the application, we briefly explain Bayesian hypothesis testing for the number of lags in VAR Since we do not know the actual lag length for the VAR and choice of the appropriate lag length affects the cointegration analysis, we apply our method that explained in Section to select the lag length Let's consider a VAR model, Y = XB + E, where B = μ 0 Φ 0 1 Φ 0 p, and X consists of vectors of lagged Y and 1s in the first column With conjugate and/or diffuse priors, we have the posteriors which are similar to our posteriors that are given in Section Then compute the Bayes factor for each Φ i =0to select the appropriate lag length Note that for this test we do not assign the correction factor C in Bayes factor The Bayes factor selects the appropriate lag length in the VAR is 1 dominantly, though the AIC indicates lags in the VAR We model a VAR(1) with a constant term for our cointegration analysis The prior specifications are the same as in the previous Monte Carlo experiments We assign equal prior probability to each rank Table illustrates the posterior probabilities with = 1 and 001 and p-values of Johansen's trace test For = 1 we see that the posterior probabilities indicate that there is one cointegration relation with 985 per cent For =0:01 our method supports one cointegration relation more strongly with 99 per cent Thus, both results by different prior specification select one cointegration relation, although Johansen's test results in no cointegration Table : Selection of the Cointegration Rank rank Pr(r Y) w/ =1 Pr(r Y) w/ =0:01 Johansen's p-value r = r = r = r = μ- 1 0:8 0:109 The posterior means of the cointegration relation among the variables Y t = m t p t ; y t ; R t is fi = Note that the first element offi is restricted to be 1 for identification The left column of Figure 1 shows the posterior densities of the second and the third element of fi The right column of Figure 1 shows the posterior densities of the three elements in ff 1
15 Figure 1: Posterior Densities of fi? (the left column) and ff (the right column) 0 15 Posterior density of b1 b1 Posterior density of alpha1 a Posterior density of alpha a Posterior density of b b Posterior density of alpha a Conclusion This paper shows simple methods of Bayesian cointegration analysis The Bayes factors are used for computing the posterior probabilities for each rank Monte Carlo experiments show that the methods proposed in this paper provide fairly good results The method that we propose requires to impose r restrictions on fi for each possible rank and then to estimate fi for each rank before computing Bayes factors for rank selection Thus, the choice of the restrictions on fi affects the values of Bayes factors in some degree Another disadvantage of the method is computing time Computing time depends upon the algorithm we choose for estimating the cointegrating vectors In this paper the Griddy- Gibbs sampler, which requires heavy computations, is chosen simply because we do not need to assign an approximation function that is needed in Metropolis-Hastings or importance sampling However, it will not be a problem in the future with much faster computers In this paper, a matrix-variate normal density for the cointegrating vector is chosen as a prior Instead, Jeffrey's or reference priors are also worth considering as Kleibergen and Paap used Jeffreys' priors for their cointegration analysis References [1] Bauwens, L, and Giot, P 1998, A Gibbs Sampling Approach to Cointegration, Computational Statistics, 1, 9-8 [] Bauwens, L, and Lubrano, M 199, Identification Restrictions and Posterior Densities in Cointegrated Gaussian VAR Systems, in TB Fomby (ed), Advances in Econometrics, vol 11B, JAI press, Greenwich, CT 1
16 [] Bauwens, L, Lubrano, M, and Richard, J-F 1999, Bayesian Inference in Dynamic Econometric Models, Oxford University Press, Oxford [] Chib, S 1995, Marginal Likelihood From the Gibbs Output, Journal of the American Statistical Association, 90, [5] Chib, S, and Greenberg, E 1995, Understanding the Metropolis-Hastings Algorithm, The American Statistician, 9, -5 [] Chen, M-H 199, Importance-Weighted Marginal Bayesian Posterior Density Estimation, Journal of the American Statistical Association, 89, [] Chen, M-H, Shao, Q-M, and Ibrahim, J-G 000, Monte Carlo Methods in Bayesian Computation, Springer Series in Statistics, Springer [8] Dickey, J, 191, The Weighted Likelihood Ratio, Linear Hypothesis on Normal Location Parameters, The Annals of Mathematical Statistics,, 0- [9] Doornik, JA 1998, OX An Object-Oriented Matrix programming Language, Timberlake Consultants Press, London [10] Evans, M and Swartz, T 000, Approximating Integrals via Monte Carlo and Deterministic Methods, Oxford University Press, Oxford [11] Geweke, J 1989, Bayesian inference in Econometric Models Using Monte Carlo Integration, Econometrica, 5, [1] Geweke, J 199, Bayesian Reduced Rank Regression in Econometrics, Journal of Econometrics,5, 1-1 [1] Goldfeld, S, and Sichel, D, 1990, The Demand for Money, Chapter 8 in B Friedman and F Hahn (eds), Handbook of Monetary Economics, Vol I, North-Holland, New York [1] Johansen, S 1991, The Power Function for the Likelihood Ratio Test for Cointegration, in J Gruber, ed, Econometric Decision Models: New Methods of Modeling and Applications, Springer Verlag, New York, NY, -5 [15] Johansen, S 000, A Small Sample Correction of the Test for Cointegrating Rank in the Vector Autoregressive Model, EUI Working Paper, ECO No 000/15, European University Institute, Italy [1] Kass, RE, and Raftery, A E 1995, Bayes Factors, Journal of the American Statistical Association, 90, -95 [1] Kass, R E, and Wasserman, L 199, The Selection of Prior Distributions by Formal Rules, Journal of the American Statistical Association, 91, 1-10 [18] Kleibergen, F, and van Dijk, HK 199, On the Shape of the Likelihood/Posterior in Cointegration Models, Econometric Theory, 10,
17 [19] Kleibergen, F, and Paap, R 1999, Priors, Posteriors and Bayes Factors for a Bayesian Analysis of Cointegration, rewritten version of Econometric Institute Report 981/A, Erasmus University Rotterdam [0] Lucas, R, 1988, Money Demand in the United States: A Quantitative Review, Carnegie- Rochester Conference Series on Public Policy Vol 9, 1-18 [1] Newton, MA, and Raftery, A E, 199, Approximate Bayesian Inference by the Weighted Likelihood Bootstrap, Journal of the Royal Statistical Society, Ser B, 5, -8 [] Phillips, P C B 199, Econometric Model Determination, Econometrica,, -81 [] Phillips, P C B, and Ploberger, W 199, An Asymptotic Theory of Bayesian Inference for Time Series, Econometrica,, 81-1 [] Ritter C, and Tanner, MA 199, Facilitating the Gibbs Sampler: The Gibbs Stopper and the Griddy-Gibbs Sampler, Journal of the American Statistical Association, 8, [5] Tierney, L, and Kadane, JB 198, Accurate Approximations for Posterior Moments and Marginal Densities, Journal of the American Statistical Association, 81, 8-8 [] Tsurumi H, and Wago, H 199, A Bayesian Analysis of Unit Root and Cointegration with an Application to a Yen-Dollar Exchange Rate Model, in TB Fomby (ed), Advances in Econometrics, vol 11B, JAI press, Greenwich, CT [] Verdinelli, I, and Wasserman, L 1995, Computing Bayes Factors Using a Generalization of the Savage-Dickey Density Ratio, Journal of the American Statistical Association, 90, 1-18 [8] Zellner, A 198, On Assessing Prior Distributions and Bayesian Regression Analysis with g-prior Distributions, ed PK Goel and A Zellner, Bayesian Inference anddecision Techniques: Essays in Honor of Bruno de Finetti, North-Holland, Amsterdam, - 1
Katsuhiro Sugita Faculty of Law and Letters, University of the Ryukyus. Abstract
Bayesian analysis of a vector autoregressive model with multiple structural breaks Katsuhiro Sugita Faculty of Law and Letters, University of the Ryukyus Abstract This paper develops a Bayesian approach
More informationBayesian Approaches to Cointegration
Bayesian Approaches to Cointegration Gary Koop Department of Economics University of Leicester Gary.Koop@leicester.ac.uk Rodney Strachan Department of Economics University of Keele r.w.strachan@econ.keele.ac.uk
More informationWorking Papers in Econometrics and Applied Statistics
T h e U n i v e r s i t y o f NEW ENGLAND Working Papers in Econometrics and Applied Statistics Finite Sample Inference in the SUR Model Duangkamon Chotikapanich and William E. Griffiths No. 03 - April
More informationThe Bayesian Approach to Multi-equation Econometric Model Estimation
Journal of Statistical and Econometric Methods, vol.3, no.1, 2014, 85-96 ISSN: 2241-0384 (print), 2241-0376 (online) Scienpress Ltd, 2014 The Bayesian Approach to Multi-equation Econometric Model Estimation
More informationLabor-Supply Shifts and Economic Fluctuations. Technical Appendix
Labor-Supply Shifts and Economic Fluctuations Technical Appendix Yongsung Chang Department of Economics University of Pennsylvania Frank Schorfheide Department of Economics University of Pennsylvania January
More informationBootstrapping the Grainger Causality Test With Integrated Data
Bootstrapping the Grainger Causality Test With Integrated Data Richard Ti n University of Reading July 26, 2006 Abstract A Monte-carlo experiment is conducted to investigate the small sample performance
More informationLecture Notes based on Koop (2003) Bayesian Econometrics
Lecture Notes based on Koop (2003) Bayesian Econometrics A.Colin Cameron University of California - Davis November 15, 2005 1. CH.1: Introduction The concepts below are the essential concepts used throughout
More informationModel Selection and Adaptive Markov chain Monte Carlo for Bayesian Cointegrated VAR model
Model Selection and Adaptive Markov chain Monte Carlo for Bayesian Cointegrated VAR model 15 th International Conference on Computing in Economics and Finance. Gareth W. Peters CSIRO Mathematical and Information
More information7. Estimation and hypothesis testing. Objective. Recommended reading
7. Estimation and hypothesis testing Objective In this chapter, we show how the election of estimators can be represented as a decision problem. Secondly, we consider the problem of hypothesis testing
More informationApproximating Bayesian Posterior Means Using Multivariate Gaussian Quadrature
Approximating Bayesian Posterior Means Using Multivariate Gaussian Quadrature John A.L. Cranfield Paul V. Preckel Songquan Liu Presented at Western Agricultural Economics Association 1997 Annual Meeting
More informationMarkov Chain Monte Carlo Methods
Markov Chain Monte Carlo Methods John Geweke University of Iowa, USA 2005 Institute on Computational Economics University of Chicago - Argonne National Laboaratories July 22, 2005 The problem p (θ, ω I)
More informationKazuhiko Kakamu Department of Economics Finance, Institute for Advanced Studies. Abstract
Bayesian Estimation of A Distance Functional Weight Matrix Model Kazuhiko Kakamu Department of Economics Finance, Institute for Advanced Studies Abstract This paper considers the distance functional weight
More informationA note on Reversible Jump Markov Chain Monte Carlo
A note on Reversible Jump Markov Chain Monte Carlo Hedibert Freitas Lopes Graduate School of Business The University of Chicago 5807 South Woodlawn Avenue Chicago, Illinois 60637 February, 1st 2006 1 Introduction
More informationA Test of Cointegration Rank Based Title Component Analysis.
A Test of Cointegration Rank Based Title Component Analysis Author(s) Chigira, Hiroaki Citation Issue 2006-01 Date Type Technical Report Text Version publisher URL http://hdl.handle.net/10086/13683 Right
More informationCointegrated VAR s. Eduardo Rossi University of Pavia. November Rossi Cointegrated VAR s Financial Econometrics / 56
Cointegrated VAR s Eduardo Rossi University of Pavia November 2013 Rossi Cointegrated VAR s Financial Econometrics - 2013 1 / 56 VAR y t = (y 1t,..., y nt ) is (n 1) vector. y t VAR(p): Φ(L)y t = ɛ t The
More informationEfficient Posterior Simulation for Cointegrated Models with Priors On the Cointegration Space
Efficient Posterior Simulation for Cointegrated Models with Priors On the Cointegration Space Gary Koop Roberto León-González Rodney W. Strachan ABSTRACT: A message coming out of the recent Bayesian literature
More informationHypothesis Testing. Econ 690. Purdue University. Justin L. Tobias (Purdue) Testing 1 / 33
Hypothesis Testing Econ 690 Purdue University Justin L. Tobias (Purdue) Testing 1 / 33 Outline 1 Basic Testing Framework 2 Testing with HPD intervals 3 Example 4 Savage Dickey Density Ratio 5 Bartlett
More informationNumerically Stable Cointegration Analysis
Numerically Stable Cointegration Analysis Jurgen A. Doornik Nuffield College, University of Oxford, Oxford OX1 1NF R.J. O Brien Department of Economics University of Southampton November 3, 2001 Abstract
More informationDynamic probabilities of restrictions in state space models: An application to the New Keynesian Phillips Curve
Dynamic probabilities of restrictions in state space models: An application to the New Keynesian Phillips Curve Gary Koop Department of Economics University of Strathclyde Scotland Gary.Koop@strath.ac.uk
More informationIndex. Pagenumbersfollowedbyf indicate figures; pagenumbersfollowedbyt indicate tables.
Index Pagenumbersfollowedbyf indicate figures; pagenumbersfollowedbyt indicate tables. Adaptive rejection metropolis sampling (ARMS), 98 Adaptive shrinkage, 132 Advanced Photo System (APS), 255 Aggregation
More informationRegime-Switching Cointegration
Regime-Switching Cointegration Markus Jochmann Newcastle University Rimini Centre for Economic Analysis Gary Koop University of Strathclyde Rimini Centre for Economic Analysis May 2011 Abstract We develop
More informationDynamic Factor Models and Factor Augmented Vector Autoregressions. Lawrence J. Christiano
Dynamic Factor Models and Factor Augmented Vector Autoregressions Lawrence J Christiano Dynamic Factor Models and Factor Augmented Vector Autoregressions Problem: the time series dimension of data is relatively
More informationCointegration Lecture I: Introduction
1 Cointegration Lecture I: Introduction Julia Giese Nuffield College julia.giese@economics.ox.ac.uk Hilary Term 2008 2 Outline Introduction Estimation of unrestricted VAR Non-stationarity Deterministic
More informationSession 3A: Markov chain Monte Carlo (MCMC)
Session 3A: Markov chain Monte Carlo (MCMC) John Geweke Bayesian Econometrics and its Applications August 15, 2012 ohn Geweke Bayesian Econometrics and its Session Applications 3A: Markov () chain Monte
More informationAccept-Reject Metropolis-Hastings Sampling and Marginal Likelihood Estimation
Accept-Reject Metropolis-Hastings Sampling and Marginal Likelihood Estimation Siddhartha Chib John M. Olin School of Business, Washington University, Campus Box 1133, 1 Brookings Drive, St. Louis, MO 63130.
More informationKobe University Repository : Kernel
Kobe University Repository : Kernel タイトル Title 著者 Author(s) 掲載誌 巻号 ページ Citation 刊行日 Issue date 資源タイプ Resource Type 版区分 Resource Version 権利 Rights DOI URL Note on the Sampling Distribution for the Metropolis-
More informationParameter estimation and forecasting. Cristiano Porciani AIfA, Uni-Bonn
Parameter estimation and forecasting Cristiano Porciani AIfA, Uni-Bonn Questions? C. Porciani Estimation & forecasting 2 Temperature fluctuations Variance at multipole l (angle ~180o/l) C. Porciani Estimation
More informationSubjective and Objective Bayesian Statistics
Subjective and Objective Bayesian Statistics Principles, Models, and Applications Second Edition S. JAMES PRESS with contributions by SIDDHARTHA CHIB MERLISE CLYDE GEORGE WOODWORTH ALAN ZASLAVSKY \WILEY-
More informationMarginal Specifications and a Gaussian Copula Estimation
Marginal Specifications and a Gaussian Copula Estimation Kazim Azam Abstract Multivariate analysis involving random variables of different type like count, continuous or mixture of both is frequently required
More informationMultivariate Time Series: Part 4
Multivariate Time Series: Part 4 Cointegration Gerald P. Dwyer Clemson University March 2016 Outline 1 Multivariate Time Series: Part 4 Cointegration Engle-Granger Test for Cointegration Johansen Test
More informationVector Autoregressive Model. Vector Autoregressions II. Estimation of Vector Autoregressions II. Estimation of Vector Autoregressions I.
Vector Autoregressive Model Vector Autoregressions II Empirical Macroeconomics - Lect 2 Dr. Ana Beatriz Galvao Queen Mary University of London January 2012 A VAR(p) model of the m 1 vector of time series
More informationBayesian Prediction of Code Output. ASA Albuquerque Chapter Short Course October 2014
Bayesian Prediction of Code Output ASA Albuquerque Chapter Short Course October 2014 Abstract This presentation summarizes Bayesian prediction methodology for the Gaussian process (GP) surrogate representation
More information(Y jz) t (XjZ) 0 t = S yx S yz S 1. S yx:z = T 1. etc. 2. Next solve the eigenvalue problem. js xx:z S xy:z S 1
Abstract Reduced Rank Regression The reduced rank regression model is a multivariate regression model with a coe cient matrix with reduced rank. The reduced rank regression algorithm is an estimation procedure,
More informationForecasting Levels of log Variables in Vector Autoregressions
September 24, 200 Forecasting Levels of log Variables in Vector Autoregressions Gunnar Bårdsen Department of Economics, Dragvoll, NTNU, N-749 Trondheim, NORWAY email: gunnar.bardsen@svt.ntnu.no Helmut
More informationECON 4160, Lecture 11 and 12
ECON 4160, 2016. Lecture 11 and 12 Co-integration Ragnar Nymoen Department of Economics 9 November 2017 1 / 43 Introduction I So far we have considered: Stationary VAR ( no unit roots ) Standard inference
More informationOnline appendix to On the stability of the excess sensitivity of aggregate consumption growth in the US
Online appendix to On the stability of the excess sensitivity of aggregate consumption growth in the US Gerdie Everaert 1, Lorenzo Pozzi 2, and Ruben Schoonackers 3 1 Ghent University & SHERPPA 2 Erasmus
More informationIntroduction to Eco n o m et rics
2008 AGI-Information Management Consultants May be used for personal purporses only or by libraries associated to dandelon.com network. Introduction to Eco n o m et rics Third Edition G.S. Maddala Formerly
More informationG. S. Maddala Kajal Lahiri. WILEY A John Wiley and Sons, Ltd., Publication
G. S. Maddala Kajal Lahiri WILEY A John Wiley and Sons, Ltd., Publication TEMT Foreword Preface to the Fourth Edition xvii xix Part I Introduction and the Linear Regression Model 1 CHAPTER 1 What is Econometrics?
More informationBayesian Analysis of AR (1) model
Abstract: Bayesian Analysis of AR (1) model Hossein Masoumi Karakani, University of Pretoria, South Africa Janet van Niekerk, University of Pretoria, South Africa Paul van Staden, University of Pretoria,
More informationEstimating and Identifying Vector Autoregressions Under Diagonality and Block Exogeneity Restrictions
Estimating and Identifying Vector Autoregressions Under Diagonality and Block Exogeneity Restrictions William D. Lastrapes Department of Economics Terry College of Business University of Georgia Athens,
More informationNonlinear Inequality Constrained Ridge Regression Estimator
The International Conference on Trends and Perspectives in Linear Statistical Inference (LinStat2014) 24 28 August 2014 Linköping, Sweden Nonlinear Inequality Constrained Ridge Regression Estimator Dr.
More informationTitle. Description. var intro Introduction to vector autoregressive models
Title var intro Introduction to vector autoregressive models Description Stata has a suite of commands for fitting, forecasting, interpreting, and performing inference on vector autoregressive (VAR) models
More informationSheffield Economic Research Paper Series. SERP Number:
Sheffield Economic Research Paper Series SERP Number: 2004013 Jamie Gascoigne Estimating threshold vector error-correction models with multiple cointegrating relationships. November 2004 * Corresponding
More informationBayesian Heteroskedasticity-Robust Regression. Richard Startz * revised February Abstract
Bayesian Heteroskedasticity-Robust Regression Richard Startz * revised February 2015 Abstract I offer here a method for Bayesian heteroskedasticity-robust regression. The Bayesian version is derived by
More informationPOSTERIOR ANALYSIS OF THE MULTIPLICATIVE HETEROSCEDASTICITY MODEL
COMMUN. STATIST. THEORY METH., 30(5), 855 874 (2001) POSTERIOR ANALYSIS OF THE MULTIPLICATIVE HETEROSCEDASTICITY MODEL Hisashi Tanizaki and Xingyuan Zhang Faculty of Economics, Kobe University, Kobe 657-8501,
More informationGibbs Sampling in Linear Models #2
Gibbs Sampling in Linear Models #2 Econ 690 Purdue University Outline 1 Linear Regression Model with a Changepoint Example with Temperature Data 2 The Seemingly Unrelated Regressions Model 3 Gibbs sampling
More informationStock index returns density prediction using GARCH models: Frequentist or Bayesian estimation?
MPRA Munich Personal RePEc Archive Stock index returns density prediction using GARCH models: Frequentist or Bayesian estimation? Ardia, David; Lennart, Hoogerheide and Nienke, Corré aeris CAPITAL AG,
More informationσ(a) = a N (x; 0, 1 2 ) dx. σ(a) = Φ(a) =
Until now we have always worked with likelihoods and prior distributions that were conjugate to each other, allowing the computation of the posterior distribution to be done in closed form. Unfortunately,
More informationA Comparison of Bayesian Model Selection based on MCMC with an application to GARCH-Type Models
A Comparison of Bayesian Model Selection based on MCMC with an application to GARCH-Type Models Tatiana Miazhynskaia Sylvia Frühwirth-Schnatter Georg Dorffner Report No. 83 November, 23 November, 23 SFB
More informationForecasting with Bayesian Vector Autoregressions
WORKING PAPER 12/2012 Forecasting with Bayesian Vector Autoregressions Sune Karlsson Statistics ISSN 1403-0586 http://www.oru.se/institutioner/handelshogskolan-vid-orebro-universitet/forskning/publikationer/working-papers/
More informationY t = ΦD t + Π 1 Y t Π p Y t p + ε t, D t = deterministic terms
VAR Models and Cointegration The Granger representation theorem links cointegration to error correction models. In a series of important papers and in a marvelous textbook, Soren Johansen firmly roots
More informationECON 4160, Spring term Lecture 12
ECON 4160, Spring term 2013. Lecture 12 Non-stationarity and co-integration 2/2 Ragnar Nymoen Department of Economics 13 Nov 2013 1 / 53 Introduction I So far we have considered: Stationary VAR, with deterministic
More informationeqr094: Hierarchical MCMC for Bayesian System Reliability
eqr094: Hierarchical MCMC for Bayesian System Reliability Alyson G. Wilson Statistical Sciences Group, Los Alamos National Laboratory P.O. Box 1663, MS F600 Los Alamos, NM 87545 USA Phone: 505-667-9167
More informationOn the Error Correction Model for Functional Time Series with Unit Roots
On the Error Correction Model for Functional Time Series with Unit Roots Yoosoon Chang Department of Economics Indiana University Bo Hu Department of Economics Indiana University Joon Y. Park Department
More informationDefault Priors and Effcient Posterior Computation in Bayesian
Default Priors and Effcient Posterior Computation in Bayesian Factor Analysis January 16, 2010 Presented by Eric Wang, Duke University Background and Motivation A Brief Review of Parameter Expansion Literature
More informationForecasting 1 to h steps ahead using partial least squares
Forecasting 1 to h steps ahead using partial least squares Philip Hans Franses Econometric Institute, Erasmus University Rotterdam November 10, 2006 Econometric Institute Report 2006-47 I thank Dick van
More informationEstimation and Testing for Common Cycles
Estimation and esting for Common Cycles Anders Warne February 27, 2008 Abstract: his note discusses estimation and testing for the presence of common cycles in cointegrated vector autoregressions A simple
More informationThe causal relationship between energy consumption and GDP in Turkey
The causal relationship between energy consumption and GDP in Turkey Huseyin Kalyoncu1, Ilhan Ozturk2, Muhittin Kaplan1 1Meliksah University, Faculty of Economics and Administrative Sciences, 38010, Kayseri,
More informationA Bayesian Approach to Testing for Markov Switching in Univariate and Dynamic Factor Models
A Bayesian Approach to Testing for Markov Switching in Univariate and Dynamic Factor Models by Chang-Jin Kim Korea University and Charles R. Nelson University of Washington August 25, 1998 Kim: Dept. of
More informationEstimating marginal likelihoods from the posterior draws through a geometric identity
Estimating marginal likelihoods from the posterior draws through a geometric identity Johannes Reichl Energy Institute at the Johannes Kepler University Linz E-mail for correspondence: reichl@energieinstitut-linz.at
More informationBayesian Assessment of Hypotheses and Models
8 Bayesian Assessment of Hypotheses and Models This is page 399 Printer: Opaque this 8. Introduction The three preceding chapters gave an overview of how Bayesian probability models are constructed. Once
More information1 Phelix spot and futures returns: descriptive statistics
MULTIVARIATE VOLATILITY MODELING OF ELECTRICITY FUTURES: ONLINE APPENDIX Luc Bauwens 1, Christian Hafner 2, and Diane Pierret 3 October 13, 2011 1 Phelix spot and futures returns: descriptive statistics
More informationRegime-Switching Cointegration
Regime-Switching Cointegration Markus Jochmann Newcastle University Rimini Centre for Economic Analysis Gary Koop University of Strathclyde Rimini Centre for Economic Analysis July 2013 Abstract We develop
More informationBayesian methods for categorical data under informative censoring
Bayesian Analysis (2008) 3, Number 3, pp. 541 554 Bayesian methods for categorical data under informative censoring Thomas J. Jiang and James M. Dickey Abstract. Bayesian methods are presented for categorical
More informationA Bayesian analysis of the PPP puzzle using an unobserved components model
A Bayesian analysis of the PPP puzzle using an unobserved components model Richard Kleijn Herman K. van Dijk Econometric Institute Report EI 2001-35 Abstract The failure to describe the time series behaviour
More informationOn the robustness of cointegration tests when series are fractionally integrated
On the robustness of cointegration tests when series are fractionally integrated JESUS GONZALO 1 &TAE-HWYLEE 2, 1 Universidad Carlos III de Madrid, Spain and 2 University of California, Riverside, USA
More informationBayesian Estimation of DSGE Models 1 Chapter 3: A Crash Course in Bayesian Inference
1 The views expressed in this paper are those of the authors and do not necessarily reflect the views of the Federal Reserve Board of Governors or the Federal Reserve System. Bayesian Estimation of DSGE
More informationThe Drachma/Deutschemark Exchange Rate, : A Monetary Analysis
The Drachma/Deutschemark Exchange Rate, 980-997: A Monetary Analysis by Athanasios P. Papadopoulos (University of Crete) and George Zis (Manchester Metropolitan University) Abstract: The validity of the
More informationMultivariate Versus Multinomial Probit: When are Binary Decisions Made Separately also Jointly Optimal?
Multivariate Versus Multinomial Probit: When are Binary Decisions Made Separately also Jointly Optimal? Dale J. Poirier and Deven Kapadia University of California, Irvine March 10, 2012 Abstract We provide
More informationRobust Estimation of ARMA Models with Near Root Cancellation. Timothy Cogley. Richard Startz * June Abstract
Robust Estimation of ARMA Models with Near Root Cancellation Timothy Cogley Richard Startz * June 2013 Abstract Standard estimation of ARMA models in which the AR and MA roots nearly cancel, so that individual
More informationDepartment of Agricultural Economics & Agribusiness
Department of Agricultural Economics & Agribusiness Monte Carlo Evidence on Cointegration and Causation Hector O. Zapata Department of Agricultural Economics and Agribusiness Louisiana State University
More informationCointegrated VAR s. Eduardo Rossi University of Pavia. November Rossi Cointegrated VAR s Fin. Econometrics / 31
Cointegrated VAR s Eduardo Rossi University of Pavia November 2014 Rossi Cointegrated VAR s Fin. Econometrics - 2014 1 / 31 B-N decomposition Give a scalar polynomial α(z) = α 0 + α 1 z +... + α p z p
More informationEstimating the marginal likelihood with Integrated nested Laplace approximation (INLA)
Estimating the marginal likelihood with Integrated nested Laplace approximation (INLA) arxiv:1611.01450v1 [stat.co] 4 Nov 2016 Aliaksandr Hubin Department of Mathematics, University of Oslo and Geir Storvik
More informationDoing Bayesian Integrals
ASTR509-13 Doing Bayesian Integrals The Reverend Thomas Bayes (c.1702 1761) Philosopher, theologian, mathematician Presbyterian (non-conformist) minister Tunbridge Wells, UK Elected FRS, perhaps due to
More informationA Bayesian Analysis of Unit Roots in Panel Data Models with Cross-sectional Dependence
A Bayesian Analysis of Unit Roots in Panel Data Models with Cross-sectional Dependence Loukia Meligkotsidou Elias Tzavalis Ioannis D. Vrontos Abstract In this paper a Bayesian approach to unit root testing
More informationOil price and macroeconomy in Russia. Abstract
Oil price and macroeconomy in Russia Katsuya Ito Fukuoka University Abstract In this note, using the VEC model we attempt to empirically investigate the effects of oil price and monetary shocks on the
More informationA TIME SERIES PARADOX: UNIT ROOT TESTS PERFORM POORLY WHEN DATA ARE COINTEGRATED
A TIME SERIES PARADOX: UNIT ROOT TESTS PERFORM POORLY WHEN DATA ARE COINTEGRATED by W. Robert Reed Department of Economics and Finance University of Canterbury, New Zealand Email: bob.reed@canterbury.ac.nz
More informationEC408 Topics in Applied Econometrics. B Fingleton, Dept of Economics, Strathclyde University
EC48 Topics in Applied Econometrics B Fingleton, Dept of Economics, Strathclyde University Applied Econometrics What is spurious regression? How do we check for stochastic trends? Cointegration and Error
More informationBayesian Statistical Methods. Jeff Gill. Department of Political Science, University of Florida
Bayesian Statistical Methods Jeff Gill Department of Political Science, University of Florida 234 Anderson Hall, PO Box 117325, Gainesville, FL 32611-7325 Voice: 352-392-0262x272, Fax: 352-392-8127, Email:
More informationTAKEHOME FINAL EXAM e iω e 2iω e iω e 2iω
ECO 513 Spring 2015 TAKEHOME FINAL EXAM (1) Suppose the univariate stochastic process y is ARMA(2,2) of the following form: y t = 1.6974y t 1.9604y t 2 + ε t 1.6628ε t 1 +.9216ε t 2, (1) where ε is i.i.d.
More informationBayesian IV: the normal case with multiple endogenous variables
Baesian IV: the normal case with multiple endogenous variables Timoth Cogle and Richard Startz revised Januar 05 Abstract We set out a Gibbs sampler for the linear instrumental-variable model with normal
More informationOutlier detection in ARIMA and seasonal ARIMA models by. Bayesian Information Type Criteria
Outlier detection in ARIMA and seasonal ARIMA models by Bayesian Information Type Criteria Pedro Galeano and Daniel Peña Departamento de Estadística Universidad Carlos III de Madrid 1 Introduction The
More informationTime Series Analysis. James D. Hamilton PRINCETON UNIVERSITY PRESS PRINCETON, NEW JERSEY
Time Series Analysis James D. Hamilton PRINCETON UNIVERSITY PRESS PRINCETON, NEW JERSEY & Contents PREFACE xiii 1 1.1. 1.2. Difference Equations First-Order Difference Equations 1 /?th-order Difference
More informationBayes methods for categorical data. April 25, 2017
Bayes methods for categorical data April 25, 2017 Motivation for joint probability models Increasing interest in high-dimensional data in broad applications Focus may be on prediction, variable selection,
More informationBayesian inference for the mixed conditional heteroskedasticity model
Econometrics Journal (2007), volume 0, pp. 408 425. doi: 0./j.368-423X.2007.0023.x Bayesian inference for the mixed conditional heteroskedasticity model L. BAUWENS AND J.V.K. ROMBOUTS CORE and Department
More informationBayesian Inference in GLMs. Frequentists typically base inferences on MLEs, asymptotic confidence
Bayesian Inference in GLMs Frequentists typically base inferences on MLEs, asymptotic confidence limits, and log-likelihood ratio tests Bayesians base inferences on the posterior distribution of the unknowns
More informationPart 6: Multivariate Normal and Linear Models
Part 6: Multivariate Normal and Linear Models 1 Multiple measurements Up until now all of our statistical models have been univariate models models for a single measurement on each member of a sample of
More informationIntroduction to Econometrics
Introduction to Econometrics T H I R D E D I T I O N Global Edition James H. Stock Harvard University Mark W. Watson Princeton University Boston Columbus Indianapolis New York San Francisco Upper Saddle
More informationSUPPLEMENT TO MARKET ENTRY COSTS, PRODUCER HETEROGENEITY, AND EXPORT DYNAMICS (Econometrica, Vol. 75, No. 3, May 2007, )
Econometrica Supplementary Material SUPPLEMENT TO MARKET ENTRY COSTS, PRODUCER HETEROGENEITY, AND EXPORT DYNAMICS (Econometrica, Vol. 75, No. 3, May 2007, 653 710) BY SANGHAMITRA DAS, MARK ROBERTS, AND
More informationStationarity and cointegration tests: Comparison of Engle - Granger and Johansen methodologies
MPRA Munich Personal RePEc Archive Stationarity and cointegration tests: Comparison of Engle - Granger and Johansen methodologies Faik Bilgili Erciyes University, Faculty of Economics and Administrative
More informationContents. Part I: Fundamentals of Bayesian Inference 1
Contents Preface xiii Part I: Fundamentals of Bayesian Inference 1 1 Probability and inference 3 1.1 The three steps of Bayesian data analysis 3 1.2 General notation for statistical inference 4 1.3 Bayesian
More informationLimited Dependent Variable Panel Data Models: A Bayesian. Approach
Limited Dependent Variable Panel Data Models: A Bayesian Approach Giuseppe Bruno 1 June 2005 Preliminary Draft 1 Bank of Italy Research Department. e-mail: giuseppe.bruno@bancaditalia.it 1 Introduction
More informationPractical Bayesian Quantile Regression. Keming Yu University of Plymouth, UK
Practical Bayesian Quantile Regression Keming Yu University of Plymouth, UK (kyu@plymouth.ac.uk) A brief summary of some recent work of us (Keming Yu, Rana Moyeed and Julian Stander). Summary We develops
More informationAn Encompassing Prior Generalization of the Savage-Dickey Density Ratio Test
An Encompassing Prior Generalization of the Savage-Dickey Density Ratio Test Ruud Wetzels a, Raoul P.P.P. Grasman a, Eric-Jan Wagenmakers a a University of Amsterdam, Psychological Methods Group, Roetersstraat
More informationSUPPLEMENTARY TABLES FOR: THE LIKELIHOOD RATIO TEST FOR COINTEGRATION RANKS IN THE I(2) MODEL
SUPPLEMENTARY TABLES FOR: THE LIKELIHOOD RATIO TEST FOR COINTEGRATION RANKS IN THE I(2) MODEL Heino Bohn Nielsen and Anders Rahbek Department of Economics University of Copenhagen. heino.bohn.nielsen@econ.ku.dk
More informationTime Series Analysis. James D. Hamilton PRINCETON UNIVERSITY PRESS PRINCETON, NEW JERSEY
Time Series Analysis James D. Hamilton PRINCETON UNIVERSITY PRESS PRINCETON, NEW JERSEY PREFACE xiii 1 Difference Equations 1.1. First-Order Difference Equations 1 1.2. pth-order Difference Equations 7
More informationBayesian phylogenetics. the one true tree? Bayesian phylogenetics
Bayesian phylogenetics the one true tree? the methods we ve learned so far try to get a single tree that best describes the data however, they admit that they don t search everywhere, and that it is difficult
More information1 Regression with Time Series Variables
1 Regression with Time Series Variables With time series regression, Y might not only depend on X, but also lags of Y and lags of X Autoregressive Distributed lag (or ADL(p; q)) model has these features:
More informationBayesian Regression Linear and Logistic Regression
When we want more than point estimates Bayesian Regression Linear and Logistic Regression Nicole Beckage Ordinary Least Squares Regression and Lasso Regression return only point estimates But what if we
More informationVAR Models and Cointegration 1
VAR Models and Cointegration 1 Sebastian Fossati University of Alberta 1 These slides are based on Eric Zivot s time series notes available at: http://faculty.washington.edu/ezivot The Cointegrated VAR
More information