Dear Stata users,
I'm trying to simulate (not to estimate) a bivariate Panel VAR(p) with Random Effects. The variables are prices and quantities which are taken from a specific group of goods (eg wrist watches). The panel should account for the heterogeneity of prices and quantities within the group of goods and the VAR should capture the relationship between prices and quantities. If anyone has comments on this, I would very much appreciate them.
What I'm looking for is a procedure to simulate the disturbance terms (the model itself is a completely different problem), where v_gi denotes the random effect of the i-th cross section in the g-th equation (g = p, q) and u_git the random error in period t. I assume a rather simple covariance structure, ie neither serial nor cross correlation of the u's and v's; both are independent and homoskedastic.
Let
(1) w_git = u_git + v_gi,
then
(2) Var(w_gi) is a T x T matrix with Var(u_g) + Var(v_g) on the diagonal and Var(v_g) off the diagonal, and
(3) Cov(w_pi,w_qi) is a diagonal matrix with elements Cov(u_p,u_q).
(4) Var(w) is equal to [I # Var(w_pi), I # Cov(w_pi,w_qi), I # Var(w_qi)], where I is the n x n identity matrix and # denotes the Kronecker product.
Var(w) is 2nT x 2nT, so I'm having trouble with large simulations (T = 100, n = 1,000) as Stata's limitation on matsize is 11,000, at least if there is no other possibility to simulate these random numbers from the multivariate normal than the Cholesky decomposition (-drawnorm- wasn't really helpful in a Panel VAR framework).
The result should be two series - w_1it and w_2it - with nT observations. The only thing I have is Var(w). How could I simulate these two series given this covariance matrix? (Var(w) and the code I have written so far is below - is X = vec(wp wq) distributed as desired?)
Many thanks in advance.
symmetric var_w[8,8]
c1: c1: c2: c2: c1: c1: c2: c2:
c1 c2 c1 c2 c1 c2 c1 c2
r1:r1 30
r1:r2 20 30
r2:r1 0 0 30
r2:r2 0 0 20 30
r1:r1 -5 0 0 0 70
r1:r2 0 -5 0 0 40 70
r2:r1 0 0 -5 0 0 0 70
r2:r2 0 0 0 -5 0 0 40 70
clear
set matsize 800
set more off
scalar drop _all
matrix drop _all
scalar n = 2
scalar T = 2
scalar var_up = 10
scalar var_vp = 20
scalar var_uq = 30
scalar var_vq = 40
scalar cov_upuq = -5
matrix var_wp = I(n) # (var_up * I(T) + J(T,T,var_vp))
matrix var_wq = I(n) # (var_uq * I(T) + J(T,T,var_vq))
matrix cov_wpwq = I(n) # (cov_upuq * I(T))
matrix var_w = (var_wp, cov_wpwq\ cov_wpwq, var_wq)
matrix A = cholesky(var_w)
matrix N = matuniform(2*n*T,1)
local i = 1
while `i' < 2*n*T {
matrix N[`i',1] = invnormal(N[`i',1])
local i = `i' +1
}
matrix X = A * N
matrix wp = X[1..n*T, 1]
quietly svmat wp
rename wp1 wp
matrix wq = X[n*T+1..2*n*T, 1]
svmat wq
rename wq1 wq
egen crossid = fill(1 1 2 2)
egen dateid = fill(1 2 1 2)
xtset crossid dateid
______________________________________________________________________________
Jetzt neu! Im riesigen WEB.DE Club SmartDrive Dateien freigeben und mit
Freunden teilen! http://www.freemail.web.de/club/smartdrive_ttc.htm/?mc=021134
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
