Skip to content

Flexible inputs fPortfolio possible?

4 messages · R@Nabble, Yohan Chalabi

Original
R@Nabble
# 
Hi,

Is it possible yet to allow pre-specified mean-return and/or covar-matrices
in fPortfolio? If so, where can I find instructions to achieve this? I
remember reading somewhere that this would eventually become available.

Thx,

R at N
Yohan Chalabi
# 
R> Hi,
   R> 
   R> Is it possible yet to allow pre-specified mean-return and/or
   R> covar-matrices
   R> in fPortfolio? If so, where can I find instructions to
   R> achieve this? I
   R> remember reading somewhere that this would eventually become
   R> available.
   R> 
   R> Thx,
   R> 
   R> R at N

Do you want to define your own estimators of the covariance matrix?

You can do it with the dev-version of fPortfolio available at r-forge.

a quick example :

##################
library(fPortfolio)
# only with development version of fPortfolio available on R-Forge

# now you can define your own estimator which must returns a list with a
# named list, with at least the following two entries '\$mu' and
# '\$Sigma', which represent estimators for the mean and covariance,
# respectively.
myEstimator <- 
    function(x, spec = NULL, ...) list(mu = colMeans(x), Sigma = cov(x))

Spec <- portfolioSpec() # default portfolio specification  
setEstimator(Spec) <- "myEstimator" # new estimator
Spec

# Load Data and Convert to timeSeries Object:
Data = as.timeSeries(data(smallcap.ts))
Data = Data[, c("BKE", "GG", "GYMB", "KRON")]
Data

## Compute properties of Efficient Portfolio
frontier <- portfolioFrontier(Data, Spec, "LongOnly")
plot(frontier)

####################

hope this helps,
Yohan
1 day later
R@Nabble
# 
Following your advice,  I did the following:
 
PropEstimates<-function(x,Estmu=colMeans(x),Covar=cov(x)) list(mu=Estmu,
Sigma=Covar)
####### Efficient Frontier
Data <- as.timeSeries(MatR)
NAss=ncol(Data)
NAss
SpecDef <- portfolioSpec()
Rfree = 0.04/12 
setRiskFreeRate(SpecDef)<- Rfree
TotUni<-PropEstimates(Data,colMeans(Data),cov(Data))
TotUni 
This code results in the following (what looks like correct) results:
$mu
      . . . . 
$Sigma
      . . . .
 
However, if I then input this as follows:
 
setEstimator(SpecDef)<-TotUni
ConstrLO = "LongOnly"
frontierLO = portfolioFrontier(Data, SpecDef, ConstrLO)
 
I get the following error:
 
ERROR:  
  c("'structure(list(mu = structure(c(0.0025, 0.00369565217391304, ' is not
a function, character or symbol", "'0.00271739130434783,
0.00347826086956522, 0.00858695652173913, ' is not a function, character or
symbol", "'0.00119565217391304, 0.00141304347826087, -0.00282608695652174, '
is not a function, character or symbol", "'0.0101086956521739,
0.00315217391304348, 0.0197826086956522, ' is not a function, character or
symbol", "'0.0141304347826087), .Names = c(\"FTSE.All.Share\",
\"FTSE.All.Stock.Gilts\", ' is not a function, character or symbol", 

Trying to correct it by literally applying your code, and using:

PropEstimates<-function(x, spec = NULL, ...) list(mu = colMeans(x), Sigma =
cov(x))
TotUni<-PropEstimates(Data,colMeans(Data),cov(Data))
setEstimator(SpecDef)<-"TotUni" 

I get this error:
 
ERROR:  
  variable "TotUni" of mode "function" was not found

In fact, even the "old" way of doing things, for example by leaving out any
setEstimator, results in:
 
ERROR:  
  variable "covEstimator" of mode "function" was not found
 
Is this because I'm still using 2.6.2 ???

Thx for any further insights/guidance,
 
R at N
Yohan Chalabi wrote:

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      Yohan Chalabi
    

    

      
      #
    

  


  

          

      
      
      
    

        
R> Following your advice,  I did the following:
   R> 
   R> PropEstimates<-function(x,Estmu=colMeans(x),Covar=cov(x))
   R> list(mu=Estmu,
   R> Sigma=Covar)
   R> ####### Efficient Frontier
   R> Data <- as.timeSeries(MatR)
   R> NAss=ncol(Data)
   R> NAss
   R> SpecDef <- portfolioSpec()
   R> Rfree = 0.04/12
   R> setRiskFreeRate(SpecDef)<- Rfree
   R> TotUni<-PropEstimates(Data,colMeans(Data),cov(Data))
   R> TotUni
   R> This code results in the following (what looks like correct)
   R> results:

how do you expect this code to work?? I did not recommend you to do
that!

you should specify the name of your estimator function with
"setEstimator<-".

this means, 

setEstimator(SpecDef) <- "PropEstimates" 

in your example.

Please use the code I already posted with the
*dev-version* of fPortfolio available at *R-forge*.

Yohan