sandbox/RFinance2014/data_analysis.R

In R-Finance/PortfolioAnalytics: Portfolio Analysis, including Numerical Methods for Optimization of Portfolios

library(PerformanceAnalytics)

source("data_prep.R")

figures.dir <- "figures"

# mix of blue, green, and red hues
my_colors <- c("#a6cee3", "#1f78b4", "#b2df8a", "#33a02c", "#fb9a99", "#e31a1c")

##### Equity Data for Example 1 and Example 2 #####
colors <- c(rep(my_colors[1], 15), rep(my_colors[2], 15), rep(my_colors[3], 5))
sd.order <- order(StdDev(equity.data))

# boxplot to compare return distributions
# mar(bottom, left, top, right)
# default: par(mar=c(5, 4, 4, 2) + 0.1)
png(paste(figures.dir, "equity_box.png", sep="/"), height = 500, width = 1000)
boxplot(coredata(equity.data[,sd.order]),
        cex.axis=0.8, las=3, ylab="Returns", pch=18,
        col=colors[sd.order],
        main="Return Distribution\n(sorted by StdDev)")
legend("topleft", legend=c("Large Cap", "Mid Cap", "Small Cap"), 
       fill=c(my_colors[1], my_colors[2], my_colors[3]), bty="n", cex=0.8)
dev.off()

##### edhec Data for Example 3 and Example 4 #####
p <- 0.95

png(paste(figures.dir, "relative_barvar.png", sep="/"))
charts.BarVaR(R[,1:3], width=60, methods=c("ModifiedES", "ModifiedVaR"), 
              main="Relative Value", colorset=rep(my_colors[2], 3))
dev.off()

png(paste(figures.dir, "directional_barvar.png", sep="/"))
charts.BarVaR(R[,4:6], width=60, methods=c("ModifiedES", "ModifiedVaR"), 
              main="Directional", colorset=rep(my_colors[4], 3))
dev.off()


colors <- c(rep(my_colors[2], 3), rep(my_colors[4], 3))
ES.order <- order(ES(R, p=p, invert=FALSE))

png(paste(figures.dir, "edhec_box.png", sep="/"), height = 500, width = 1000)
boxplot(coredata(R[,ES.order]),
        cex.axis=0.8, las=3, ylab="Returns", pch=18,
        col=colors[ES.order],
        main="Return Distribution\n(sorted by Modified ES (95%))")
legend("topleft", legend=c("Relative Value", "Directional"), 
       fill=c(my_colors[1], my_colors[2]), bty="n", cex=0.8)
dev.off()

# script for data analysis

# library(PerformanceAnalytics)
# library(lattice)
# library(corrplot)

# load("data/edhec.rda")
# 
# head(edhec)
# R <- edhec[,1:4]
# p <- 0.95
# 
# first(R)
# last(R)

# plot the timeseries of returns
# plot(as.zoo(edhec))
# xyplot(R, scales=list(y="same"))
# charts.BarVaR(R, width=36, methods=c("ModifiedES", "ModifiedVaR"))
# dev.off()

# boxplot to compare return distributions
# mar(bottom, left, top, right)
# default: par(mar=c(5, 4, 4, 2) + 0.1)
# par(mar=c(10, 4, 4, 2) + 0.1)
# boxplot(coredata(R[,order(ES(R, p=p, invert=FALSE))]),
#         cex.axis=0.8, las=3, ylab="Returns", pch=18,
#         main="Return Distribution\n(sorted by Modified ES (95%))")
# par(mar=c(5, 4, 4, 2) + 0.1)
# dev.off()

# head(R[,order(ES(R, invert=FALSE))])
# head(R[,order(StdDev(R))])
# chart.Boxplot(R[,order(ES(R, invert=FALSE))])
# chart.Boxplot(R[,order(StdDev(R))])
# boxplot(coredata(R), col=c(2:5), cex.names=0.8, las=3)

# chart the distribution of returns
# for(i in 1:ncol(R)){
#   chart.Histogram(R[,i], methods=c("add.density", "add.normal"), 
#                   colorset=c("lightgray", "black", "blue"))
#   legend("topleft", legend=c("kernel density estimate", "normal"), 
#          lty=c(1,1), col=c("black", "blue"), bty="n")
#   Sys.sleep(1)
# }


# chart the correlation and covariance
# from http://cran.r-project.org/web/packages/corrplot/vignettes/corrplot-intro.html
# cor.mtest <- function(mat, conf.level = 0.95) {
#   mat <- as.matrix(mat)
#   n <- ncol(mat)
#   p.mat <- lowCI.mat <- uppCI.mat <- matrix(NA, n, n)
#   diag(p.mat) <- 0
#   diag(lowCI.mat) <- diag(uppCI.mat) <- 1
#   for (i in 1:(n - 1)) {
#     for (j in (i + 1):n) {
#       tmp <- cor.test(mat[, i], mat[, j], conf.level = conf.level)
#       p.mat[i, j] <- p.mat[j, i] <- tmp$p.value
#       lowCI.mat[i, j] <- lowCI.mat[j, i] <- tmp$conf.int[1]
#       uppCI.mat[i, j] <- uppCI.mat[j, i] <- tmp$conf.int[2]
#     }
#   }
#   return(list(p.mat, lowCI.mat, uppCI.mat))
# }
# res <- cor.mtest(R)
# 
# corrplot(cor(R), p.mat=res[[1]], main="Correlation", 
#          sig.level=0.05, tl.cex=0.8)

# corrplot(M, method="number", bg="gray", tl.cex=0.8)
# corrplot.mixed(M, bg="gray", tl.cex=0.8)

# If I compare sample min variance portfolio to a ledoit-shrinkage or robust, 
# I should use plotcov to compare covaiance matrices