[R-SIG-Finance] Using forecasted data in fportfolio

[matt at considine.net]

Here's how I approached it at one point when I wanted to use a 
consultant's data as input (code is pulled from a larger script, so 
doesn't stand alone ...).  There may be much easier/better ways.

expRet # this is the expected return of the asset classes
covM   # is a positive definite or near p.d. variance/covariance matrix
	# derived from asset class correlations and standard deviations
library(timeSeries)
library(MASS)
#create returns series
testdata<- 
timeSeries(mvrnorm(n=500,mu=expRet,Sigma=covM,empirical=TRUE))
#check : this should be a set of zeros
cor(testdata)-ia2$correlation
#create price series - used elsewhere
testprices<-100*apply(testdata,2,function(x) cumprod(1+x))
testprices<-rbind(matrix(rep(100,length(expRet)),nrow=1),testprices)
colnames(testprices)<-colnames(testdata)
#check that expected means are correct
colMeans(apply(testprices,2,function(x) 
ROC(x,type="discrete")),na.rm=TRUE)-expRet
library(fPortfolioAdvanced)
the_hull<-feasleHull(testdata)
...

Hope this helps,
Matt

> List-Post: <mailto:r-sig-finance at r-project.org>
> Content-Transfer-Encoding: quoted-printable
> From: ericstrom <ericstrom at aol.com>
> Precedence: list
> MIME-Version: 1.0
> To: r-sig-finance at r-project.org
> Date: Thu, 7 Nov 2013 19:42:54 -0500 (EST)
> Message-ID: <8D0AA32987CA530-1988-16EC8 at webmail-vm004.sysops.aol.com>
> Content-Type: text/plain
> Subject: [R-SIG-Finance] Using forecasted data in fportfolio
> Message: 1
>
>  Still trying to figure out the correct way in the fportfolio package 
> to de=
> termine the portfolio frontier using a forcasted returns vector 
> and/or a fo=
> recasted covariance matrix ?  I have no problem coming up with the 
> portfoli=
> o frontier using historical returns. But what I want to do is use 
> forcasted=
>  returns and / or forecasted covariance matrix. For example, if my 
> historic=
> al data is in zts, then the following works:spec 
> =3DportfolioSpec()frontier=
>  <- portfolioFrontier(zts, spec)frontierPlot(frontier, 
> risk=3D"Sigma", a=
> uto=3DFALSE)class(zts)[1] "timeSeries"attr(,"package")[1] 
> "timeSeries"head(=
> zts)GMT                  WBLRX        LSBRX        DBLTX         
> TFSHX     =
>       TBT          SPY2012-11-09  0.003013563 -0.002106003 
> 0.0000000000 -0.=
> 0036968619 -0.0312447930 -0.0112530152012-11-16 -0.007045827 
> -0.003519891 0=
> .0000000000 -0.0037105794 -0.0074161810 -0.0130282882012-11-23  
> 0.003025721=
>   0.006326910 0.0000000000  0.0018570107  0.0337662016  
> 0.0358429662012-11-=
> 30  0.001006543  0.005590511 0.0018365478  0.0027790662 -0.0096968465 
> 0.00=
> 56951462012-12-07 -0.001006543  0.005559431 0.0009170106  
> 0.0009246418  0.0=
> 008254231  0.0017955262012-12-14  0.004020106  0.005528694 
> 0.0009161705  0.=
> 0046104277  0.0134406219 -0.002155018But suppose instead of using 
> historica=
> l returns data, I want to use forecasted returns and/or forecasted 
> covarian=
> ce (call them mu1 and covmat). What are the analogous statements to 
> portfol=
> ioFrontier and frontierPlot using mu1 and covmat as inputs instead of 
> histo=
> rical returns data ?> class(mu1)[1] "numeric"> class(covmat)[1] 
> "matr=
> ix"> mu1        WBLRX         LSBRX         DBLTX         TFSHX    
>     =
>   TBT           SPY  0.0013537910  0.0010611297  0.0001845725 
> -0.0021064133=
>   0.0032720265  0.0052493188 > covmat              WBLRX         
> LSBRX  =
>        DBLTX         TFSHX           TBT           SPYWBLRX  
> 8.980255e-05  =
> 5.538752e-06  3.384955e-06 -2.753534e-06 -2.347333e-05  
> 6.669676e-06LSBRX  =
> 5.538752e-06  5.017554e-05  1.297063e-05 -1.055112e-05 -8.994622e-05  
> 2.555=
> 718e-05DBLTX  3.384955e-06  1.297063e-05  1.848077e-05 -6.448218e-06 
> -5.496=
> 978e-05  1.561903e-05TFSHX -2.753534e-06 -1.055112e-05 -6.448218e-06  
> 4.277=
> 017e-05  4.471585e-05 -1.270549e-05TBT   -2.347333e-05 -8.994622e-05 
> -5.496=
> 978e-05  4.471585e-05  1.388306e-03 -1.083118e-04SPY    6.669676e-06  
> 2.555=
> 718e-05  1.561903e-05 -1.270549e-05 -1.083118e-04  2.170639e-04I've 
> checked=
>  the package documentation and I am not finding anything. Any help 
> would be=
>  appreciated.