ResampledPortfolio        package:fPortfolio        R Documentation

_R_e_s_a_m_p_l_e_d _P_o_r_t_f_o_l_i_o _F_r_o_n_t_i_e_r

_D_e_s_c_r_i_p_t_i_o_n:

     A collection and description of functions to  ... 

     The functions are:

       'frontierResampled'  ... .

_U_s_a_g_e:

     frontierResampled(x = as.matrix(read.csv("nova-updated.csv", header = FALSE)),
         expectedReturns = read.csv("expected-returns.csv"), frequency = 52,
         control = list(nFrontier = 200, returnChoice = 2, covarianceChoice = 2,
         targetChoice = 6, correlationChoice = 1), confidenceLevel = TRUE, 
         trace = TRUE)

_A_r_g_u_m_e_n_t_s:

       x: any rectangular multivariate time series object which can be
          transformed into a matrix throught the function 'as.matrix'.  

expectedReturns: expectedReturns 

frequency: frequency of time series records. An integer value, by
          default 52 which denotes weekly data records. 

 control: a control list with the following entries: ... 

confidenceLevel: a logical flag, if 'TRUE' then 95 returned, otherwise
          not. By default 'TRUE'. 

   trace: a logical if 'TRUE' then the portfolio resampling process
          will be ttraced, otherwise not. By default 'TRUE'. 

_D_e_t_a_i_l_s:

     I attach the code and some data (list of weekly returns of several
     assets and list of possible expected returns; they have to be in
     the same folder as the R script.

     To get at least a glimpse at what is going on, one should be
     familiar with the book  "Efficient Asset Management" by R. Michaud
      and the concept of resampled efficiency.

     Stripped to the bones, the basic idea is to Monte-Carlo simulate
     the returns of the assets in your portfolio (you sample a
     multivariate distribution which is consistent with the "true"
     covariance matrix C and the "true" vector of expected returns v
     you assume to describe your portfolio) and generate many efficient
     frontiers.

     Then one averages out of the simulated frontiers the weights of
     the "statistically equivalent" portfolios. By using them in
     conjunction with C and v one generates a resampled frontier which,
     compared to Markowitz classical treatment, is sub-optimal by
     definition but is numerically more robust (Markowitz is too
     sensitive to the hypothesis of the problem) and leads to more
     diversified portfolios which make investment sense.

     I exchanged some emails with Michaud discussing some
     technicalities of the method (of course up to a certain extent:
     there are two patents pending).

     A few issues need further thought, but that can be postponed for
     now. Beside this, the code implements some shrinkage techniques
     for the expected returns of the assets and the covariance matrix
     (Bayesian statistics with different priors, see the works by
     Olivier Ledoit & Michael Wolf for the formulas).

     Confidence levels coming from a re-interpretation of the research
     by Jobson & Korkie are implemented as in the book by Michaud.

_V_a_l_u_e:

     A list ...

_A_u_t_h_o_r(_s):

     Lorenzo Isella for the R code,
      Diethelm Wuertz for this Rmetrics port.

_R_e_f_e_r_e_n_c_e_s:

     Michaud, R.O. (1998);  _Efficient Asset Management: A Practical
     Guide to Stock  Portfolio Optimization and Asset Allocation _, 
     President and Fallows of the Harvard College, USA.

_E_x_a_m_p_l_e_s:

     ## SOURCE("fPortfolio.102D-ResampledPortfolio")

     ## Not run: 
     ## file -
        frontierResampled()
     ## End(Not run) 

