Dear R users,
I am trying to carry out MLE of the time-varying CAPM using the FKF package.
My approach so far has been to try and adapt the example given in the help
file found using ?fkf which demonstrates the MLE of an ARMA(2,1) model.
When I attempt to run my R code (given below) I get the following error:
Error in fkf(a0 = sp$a0, P0 = sp$P0, dt = sp$dt, ct = sp$ct, Tt = sp$Tt, :
Some of dim(dt)[2], dim(ct)[2], dim(Tt)[3],
dim(Zt)[3], dim(HHt)[3], dim(GGt)[3] or
dim(yt)[2] is/are neither equal to 1 nor equal to 'n'!
Here is the R code that generated this error
# Fitting time-varying parameter CAPM to BP stock
# let rt denote adjusted daily returns on a stock
# let rmt denote daily returns on the appropriate benchmark e.g. SP500
# rt = alphat + betat * rmt + et
# alphat = alphat_1 + n1t
# betat = betat_1 + n2t
# where et ~ N(0,H)
# (n1t,n2t) ~ N(0,Q)
#load required packages
library(tseries)
library(FKF)
# load data
FTSE100 <- get.hist.quote(instrument = "^FTSE", start = "2007-01-01", quote
= "Close", retclass = "zoo", quiet = TRUE)
BP <- get.hist.quote(instrument = "BP", start = "2007-01-01", quote =
"AdjClose", retclass = "zoo", quiet = TRUE)
# calulate continuously compounded daily returns
FTSE100.ret <- lag(log(FTSE100), k = -1) - log(FTSE100)
BP.ret <- lag(log(BP), k = -1) - log(BP)
# collect data
data <- merge(BP.ret,FTSE100.ret, all = FALSE)
index <- index(data)
# create data matrix
Z <- cbind(1, as.vector(data[,2]))
y <- as.matrix(data[,1],nrow = 1, ncol = length(y))
# format Z into an array
Zt <- array(NA,c(1,2,nrow(Z)))
for(i in 1:nrow(Z)) {
Zt[,,i] <- Z[i,]
}
# specify state-space form of the time-varying capm
capm.ss <- function(alpha,beta,sigma_e,sigma_n1,sigma_n2) {
Tt <- diag(1,2)
Zt <- Zt
ct <- matrix(0)
dt <- matrix(0,nrow = 2, ncol = 1)
GGt <- matrix(sigma_e^2)
HHt <- diag(c(sigma_n1^2,sigma_n2^2),2)
a0 <- c(0,0)
P0 <- diag(10^6,2)
return(list(a0 = a0, P0 = P0, ct = ct, dt = dt, Zt = Zt, Tt = Tt, GGt =
GGt, HHt = HHt))
}
# now define the loss function
loss.fn <- function(theta,yt) {
sp <- capm.ss(theta["alpha"], theta["beta"],
theta["sigma_e"],theta["sigma_n1"], theta["sigma_n2"])
ans <- fkf(a0 = sp$a0, P0 = sp$P0, dt = sp$dt, ct = sp$ct, Tt = sp$Tt,
Zt =
sp$Zt, HHt = sp$HHt, GGt = sp$GGt, yt = yt)
return(-ans$logLik)
}
start_val <- runif(5)
fit <- optim(start_val, loss.fn, hessian = TRUE, yt = y)
--
View this message in context:
http://r.789695.n4.nabble.com/time-varying-parameters-kalman-filter-estimation-problem-using-FKF-package-tp4208364p4208364.html
Sent from the R help mailing list archive at Nabble.com.
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
