Hey @donotdespair
compute_variance_decompositions gives a different result from the R function compute_fevd
Here is a reproducible example
library(bsvarSIGNs)
# R code for fevd
compute_fevd = function(irf, horizon = 40) {
N = dim(irf)[1]
S = dim(irf)[4]
fevd = array(NA,c(N,N,horizon,S))
for (s in 1:S){
for (i in 1:(horizon)){
for (n in 1:N){
for (nn in 1:N){
fevd[n,nn,i,s] = sum(irf[n,nn,1:i,s]^2)
}
}
fevd[,,i,s] = diag(1/apply(fevd[,,i,s],1,sum))%*%fevd[,,i,s]
}
}
fevd = 100*fevd
fevd
}
data(optimism)
zero_irf = matrix(0, nrow = 5, ncol = 5)
zero_irf[1, 1] = 1
sign_irf = array(0, dim = c(5, 5, 1))
sign_irf[2, 1, 1] = 1
specification = specify_bsvarSIGN$new(
optimism*100,
p = 4,
sign_irf = sign_irf,
zero_irf = zero_irf
)
posterior = estimate(specification, S = 100)
irf = compute_impulse_responses(posterior, horizon = 40)
fevd_R = compute_fevd(irf, horizon = 10)
fevd_cpp = compute_variance_decompositions(posterior, horizon = 10)
fevd_R[,,1,1]
fevd_cpp[,,1,1]
Seems like the R function is correct, have a look! Thanks.
Best,
Adam
RetroSearch is an open source project built by @garambo | Open a GitHub Issue
Search and Browse the WWW like it's 1997 | Search results from DuckDuckGo
HTML:
3.2
| Encoding:
UTF-8
| Version:
0.7.4