Barra Industry

rm(list=ls())  

library(quantmod)
library(factorAnalytics)

####################################################################
# Barra Industry Factor Model
####################################################################
# Read Fundamental Factor Data
stacked.df <- read.table("factorDataSet.csv",as.is=T,header=T,sep=",")
stacked.df[,1] <- as.yearmon(stacked.df[,1])
stacked.df <- stacked.df[which(stacked.df[,1]>="Jan 2008" & stacked.df[,1]<="Dec 2012"),]

# Create Industry Classification Matrix and Return Matrix
beta.mat <- table(stacked.df$TICKER,stacked.df$SECTOR)
beta.mat <- ifelse(beta.mat>0,1,0)
ret.zt <- t(ret.z)
ret.zt <- ret.zt[order(row.names(ret.zt)),]

# Fit Industry Factor Model via 2-Stage Weighted Least Squares
F.hat1 <- solve(crossprod(beta.mat))%*%t(beta.mat)%*%ret.zt
E.hat1 <- ret.zt-beta.mat%*%F.hat1
diagD.hat1 <- apply(E.hat1,1,var)
Dinv.hat1 <- diag(diagD.hat1^(-1))
H <- solve(t(beta.mat)%*%Dinv.hat1%*%beta.mat)%*%t(beta.mat)%*%Dinv.hat1
F.hat2 <- H%*%ret.zt
E.hat2 <- ret.zt-beta.mat%*%F.hat2
diagD.hat2 <- apply(E.hat2,1,var)
Dinv.hat2 <- diag(diagD.hat2^(-1))
F.hat2 <- t(F.hat2)
cov.ind <- beta.mat%*%var(F.hat2)%*%t(beta.mat)+diag(diagD.hat2)

# Fit Cross-sectional Factor Model via fitFundamentalFactorModel
rfr <- rep(coredata(returns.z$rf),each=30)
stacked.df$RETURN <- stacked.df$RETURN-rfr
cross.sect <- fitFundamentalFactorModel(stacked.df,exposure.names="SECTOR",returnsvar="RETURN",datevar="DATE",assetvar="TICKER")
par(las=2)
barplot(pca$r2,main="R-Squared Values for PCA Model",names.arg=colnames(ret.z),col=4)
dev.off()