Book a Demo!
CoCalc Logo Icon
StoreFeaturesDocsShareSupportNewsAboutPoliciesSign UpSign In
braverock
GitHub Repository: braverock/portfolioanalytics
 Path: blob/master/demo/higher_moments_boudt.R
1433 views
1
#' ---

2
#' title: "Higher Moment Estimates Demo"

3
#' author: Ross Bennett

4
#' date: "7/17/2014"

5
#' ---

6


7
#' This script demonstrates how to solve a constrained portfolio optimization 

8
#' problem using moments estimated via a factor model based on the work of

9
#' Kris Boudt.

10


11
#' Load package and data.

12
library(PortfolioAnalytics)

13
data(edhec)

14
R <- edhec[, 1:10]

15
funds <- colnames(R)

16


17
#' Construct initial portfolio with basic constraints.

18
init.portf <- portfolio.spec(assets=funds)

19
init.portf <- add.constraint(portfolio=init.portf, type="weight_sum",

20
                             min_sum=0.99, max_sum=1.01)

21
init.portf <- add.constraint(portfolio=init.portf, type="long_only")

22
init.portf <- add.objective(portfolio=init.portf, type="risk", name="ES",

23
                            arguments=list(p=0.9, clean="boudt"))

24


25
#' This is not necessary for the optimization, but demonstrates how the

26
#' moments are estimated using portfolio.moments.boudt.

27
cleanR <- Return.clean(R, "boudt")

28
fit <- statistical.factor.model(cleanR, 3)

29


30
#' Here we extract the moments.

31
sigma <- extractCovariance(fit)

32
m3 <- extractCoskewness(fit)

33
m4 <- extractCokurtosis(fit)

34


35
# Here we compute the moments with portfolio.moments.boudt and show that the

36
#' result is equivalent to fitting the factor model and extracting the moment

37
#' estimates.

38
moments.boudt <- portfolio.moments.boudt(R, init.portf, k=3)

39
all.equal(moments.boudt$sigma, sigma)

40
all.equal(moments.boudt$m3, m3)

41
all.equal(moments.boudt$m4, m4)

42


43
#' Generate set of random portfolios to use for the optimization.

44
rp <- random_portfolios(init.portf, 5000)

45


46
#' Here we run the optimization using sample estimates for the moments.

47
#' The default for momentFUN is set.portfolio.moments which computes

48
#' the sample estimates of the moments.

49
minES.lo.sample <- optimize.portfolio(R=R, portfolio=init.portf, 

50
                                      rp=rp, optimize_method="random",

51
                                      trace=TRUE)

52


53
#' Now we run the optimization with statistical factor model estimates of the 

54
#' moments.

55
minES.lo.boudt <- optimize.portfolio(R=R, portfolio=init.portf, 

56
                                     momentFUN=portfolio.moments.boudt, 

57
                                     k=3, rp=rp,

58
                                     optimize_method="random",

59
                                     trace=TRUE)

60


61


62


63