I have time-series data. I have returned the data and got the log returns. I aim to select the best fit ARMA-GARCH
model from a list of models. Instead of fitting each model separately, I want to fit these models at once and select the best among them. Hence, I make a warp function to do the job for me. For example, I would like to fit several models:
1- ARMA(1,1)-GARCH(1,1)
2- ARMA(1,2)-GARCH(1,1)
3- ARMA(1,3)-GARCH(1,1)
4- ARMA(2,1)-GARCH(1,1)
5- ARMA(2,2)-GARCH(1,1)
6- ARMA(2,3)-GARCH(1,1)
.. ..
1- ARMA(1,1)-GARCH(1,2)
2- ARMA(1,1)-GARCH(1,3)
3- ARMA(1,1)-GARCH(2,1)
and so on. That is, lets ARMA(p,q)
range from 1:6
, and GARCH(p,q)
also from 1:6
. How to do this using loop
function in R. I would like to return the smallest criteria (AIC
) among all the models). In other words, I would like my function to loop between different orders automatically.
Here is my try:
GarchWarp <- function(n,dat,d){ ## n is the order of the model. d is the dimension of the data, for example, one series, two series.
GarchWarp <- function(n,dat,d){
meanModel <- varModel <- list()
if(n > 1){
for(i in 1:n){
meanModel[i] <- list(armaOrder = c(i,i)) # ARMA
varModel <- list(model = "sGARCH", garchOrder = c(i,i)) # GARCH
uspec <- ugarchspec(varModel, mean.model = meanModel,
distribution.model = "std") # scaled Student t
fitPC <- apply(dat, 2, function(x) ugarchfit(uspec, data = x))
Recrit <- lapply(1:d, function(i) infocriteria(fitPC[[i]]))
}
}
return(Recrit)
}
My loop function did not work
Sample of my data: d=2
.
structure(c(0.00618750144876756, 0.00491561236137983, 0.00178237912109402,
0.0052134813154332, 0.00129554617030614, 0.00232162478086728,
0, 0.0320481134224817, 0.00881819387210214, 0, 0.00569585759405733,
-0.00259413863757629), class = c("xts", "zoo"), index = structure(c(1514851200,
1514937600, 1515024000, 1515110400, 1515369600, 1515456000), tzone = "UTC", tclass = c("POSIXct",
"POSIXt")), .Dim = c(6L, 2L), .Dimnames = list(NULL, c("P_Sha",
"P_Jap")))