With a lot of help from contributors to StackOverflow I have managed to put together a function to derive the weights of a 2-asset portfolio which maximises the Sharpe ratio. No short sales are allowed and the sum of weights add to 1. What I would like to do now is to constrain asset A to not being more or less than 10% from a user defined weight. As an example I would like to constrain the weight of asset A to be no less than 54% or more than 66% (i.e 60% +/- 10%). So on the below example I would end up with weights of (0.54,0.66) instead of the unsconstrained (0.243,0.7570) .I assume this can be done by tweaking bVect but I am not so sure how to go about it. Any help would be appreciated.
asset_A <- c(0.034320510,-0.001209628,0.031900161,0.023163947,-0.001872938,-0.010322489,0.006090395,-0.003270854,0.017778990,0.017204915)
asset_B <- c(0.047103261,0.055175057,0.021019816,0.020602347,0.007281368,-0.006547404,0.019155238,0.005494798,0.025429958,0.014929124)
require(quadprog)
HR_solve <- function(asset_A,asset_B) {
vol_A <- sd(asset_A)
vol_B <- sd(asset_B)
cor_AB <- cor(cbind(asset_A,asset_B),method="pearson")
ret_A_B <- as.matrix(c(mean(asset_A),mean(asset_B)))
vol_AB <- c(vol_A,vol_B)
covmat <- diag(as.vector(vol_AB))%*%cor_AB%*%diag(as.vector(vol_AB))
aMat <- cbind(rep(1,nrow(covmat)),diag(1,nrow(covmat)))
bVec <- c(1,0,0)
zeros <- array(0, dim = c(nrow(covmat),1))
minw <- solve.QP(covmat, zeros, aMat, bVec, meq = 1 ,factorized = FALSE)$solution
rp <- as.numeric(t(minw) %*% ret_A_B)
sp <- sqrt(t(minw) %*% covmat %*% minw)
wp <- t(matrix(minw))
sret <- sort(seq(t(minw) %*% ret_A_B,max(ret_A_B),length.out=100))
aMatt <- cbind(ret_A_B,aMat)
for (ri in sret[-1]){
bVect <- c(ri,bVec)
result <- tryCatch({solve.QP(covmat, zeros, aMatt, bVect, meq = 2,factorized = FALSE)},
warning = function(w){ return(NULL) } , error = function(w){ return(NULL)}, finally = {} )
if (!is.null(result)){
wp <- rbind(wp,as.vector(result$solution))
rp <-c(rp,t(as.vector(result$solution) %*% ret_A_B))
sp <- c(sp,sqrt(t(as.vector(result$solution)) %*% covmat %*% as.vector(result$solution))) }
}
HR_weights <- wp[which.max(rp/sp),]
as.matrix(HR_weights)
}
HR_solve(asset_A,asset_B)
[,1]
[1,] 0.2429662
[2,] 0.7570338
Ok I have found a way to do this... if you think there is a more elegant way please let me know...
I think you should take a look at the link below.
http://economistatlarge.com/portfolio-theory/r-optimized-portfolio/r-code-graph-efficient-frontier
I think you'll learn a lot from that. I'll post the code here, in case the link gets shut down sometime in the future.
Just change aMat and bVec: