Bayesian vector autoregressions with stochastic volatility
read more
Citations
What are the effects of monetary policy on output? Results from an agnostic identification procedure
Time Varying Structural Vector Autoregressions and Monetary Policy
Rao-Blackwellised Particle Filtering for Dynamic Bayesian Networks.
Adaptive pattern recognition based control system and method
Rao-blackwellised particle filtering for dynamic Bayesian networks
References
Autoregressive conditional heteroscedasticity with estimates of the variance of United Kingdom inflation
Aspects of multivariate statistical theory
ARCH modeling in finance: A review of the theory and empirical evidence
Bayesian inference in statistical analysis
Related Papers (5)
Drifts and volatilities: monetary policies and outcomes in the post WWII US
Frequently Asked Questions (10)
Q2. What is the simplest way to analyze the posterior?
Use as importance sampling density a tdistribution centered at B with Hessian J , whose degrees of freedom are chosen to ensure fatter tails than those of the marginal posterior T;marg(B): choose with 0 < < T + l + ml, preferably close to the upper bound.
Q3. What is the model used in this paper?
The stochastic volatility model used here is similar to Shephard (1994), whose model is a univariate, non-Bayesian and nonautoregressive special case of the model proposed here.
Q4. What is the simplest way to calculate the Bayesian posterior?
This paper introduced Bayesian vector autoregressions with stochastic volatility, deriving in closed form the Bayesian posterior, when the error precision matrix is stochas-9 tically time-varying.
Q5. What is the key for proving the validity of these updating formulas?
The key for proving the validity of these updating formulas here or in the next section is theorem 2 and its proof (see appendix B): as the unobserved shock #t occurs, one needs to do a \\change of variable" from d dht d#t to d dht+1 dzt for some suitably de ned zt.
Q6. What is the simplest way to solve the q-value problem?
Note that N 1 t can be computed numerically cheaply viaN 1t = N 1t 1 N 1 t 1XtX 0 tN 1 t 1=(X 0 tN 1 t 1Xt + ) = ;as can be veri ed directly or with rule (T8), p. 324 in Leamer (1978).
Q7. What is the generalization of the multiplication of two real numbers in equation (2)?
Equation (9) is one of two rather natural generalizations of the multiplication of two real numbers in equation (2) in order to guarantee the symmetry of the resulting matrix Ht+1.
Q8. What is the posterior of the HT+1?
Numerical methods are needed, since the posterior (11) is proportional to a Normal-Wishart distribution scaled with the function gT (B).
Q9. How many draws are the heavily weighted?
The weights can di er by orders of magnitude: examining the raw numbers shows that the draw with the largest weight receives 5.3% of the sum of all weights, the 109 most heavily weighted draws constitute 50% of the mass and the 741 draws with the highest weights make up 90%.
Q10. What is the gamma function in Muirhead (1982)?
For each t = 1; : : : ; T calculate et = Yt BtXt andNt = Nt 1 +XtX 0 t(12)5 Bt = Bt 1Nt 1 + YtX 0t N 1t(13)St = St 1 + et 1 X 0tN 1 t Xt e0t(14)gt(B) gt 1(B) j (B Bt)Nt(B Bt) 0 + St j 1=2(15)t( ; ) m(( + l + 1)=2)m(( + l)=2) m(l+ )=2 t 1( ; )(16)In equation (16), m( ) is the multivariate gamma function, de ned in Muirhead (1982), De nition 2.1.10.