Test de cointégration de deux stocks en utilisant les données de Yahoo Finance

Question

J'essaie d'effectuer un test de co-intégration sur deux stocks, en utilisant des données de Yahoo Finance. D'après ce que j'ai lu, il existe des moyens moins compliqués pour récupérer des données Yahoo. J'ai besoin de récupérer deux titres et les définir comme stk1 et stk2 ainsi que d'être en mesure d'ajuster la période de temps des données récupérées. Voici ce que j'ai jusqu'ici.

library(zoo)    
library(tseries)   

# Read the CSV files into data frames 
stk1 <- read.csv("http://ichart.finance.yahoo.com/table.csv?s=CAT&a=8&b=1&c=2009&d=12&e=31&f=2010&g=d&ignore=.csv", stringsAsFactors=F) 
stk2 <- read.csv("http://ichart.finance.yahoo.com/table.csv?s=DD&a=8&b=1&c=2009&d=12&e=31&f=2010&g=d&ignore=.csv", stringsAsFactors=F) 

# The first column contains dates. as.Date converts strings into Date objects 
stk1_dates <- as.Date(stk1[,1]) 
stk2_dates <- as.Date(stk2[,1]) 

# The seventh column contains the adjusted close. We use the zoo function to 
# create zoo objects from that data. The function takes two arguments: a 
# vector of data and a vector of dates. 
stk1 <- zoo(stk1[,7], stk1_dates) 
stk2 <- zoo(stk2[,7], stk2_dates) 

# The merge function combines two (or more) zoo objects, 
# computing either their intersection (all=FALSE) or union (all=TRUE). 
t.zoo <- merge(stk1, stk2, all=FALSE) 

# At this point, t.zoo is a zoo object with two columns: stk1 and stk2. 
# Most statistical functions expect a data frame for input, so we convert. 
t <- as.data.frame(t.zoo) 

# Tell the user what dates are spanned by the data. 
cat("Date range is", format(start(t.zoo)), "to", format(end(t.zoo)), "\n") 

m <- lm(stk1 ~ stk2 + 0, data=t) 
beta <- coef(m)[1] 
cat("Assumed hedge ratio is", beta, "\n") 

sprd <- t$stk1 - beta*t$stk2 
ht <- adf.test(sprd, alternative="stationary", k=0) 
cat("ADF p-value is", ht$p.value, "\n") 

if (ht$p.value < 0.05) { 
    cat("The spread is likely mean-reverting\n") 
} else { 
    cat("The spread is not mean-reverting.\n") 
} 

Quels outils existent pour rendre ceci plus facile et/ou plus robuste?

Answer 1

quantmod fournit une interface très agréable à Yahoo (et d'autres fournisseurs) données:

library(quantmod) 
library(tseries) 

stk1 <- getSymbols("DD", from="2009-01-01", auto.assign=FALSE) 
stk2 <- getSymbols("CAT", from="2009-01-01", auto.assign=FALSE) 

# UPDATE: Here's how I would approach the rest of the OP's example 
# Ad() is a helper function in quantmod 
pair <- merge(Ad(stk1), Ad(stk2), all=FALSE) 

cat("Date range is", format(start(pair)), "to", format(end(pair)), "\n") 

# build the formula with the instrument names 
eqn <- as.formula(paste(colnames(pair), collapse=" ~ 0 + ")) 
# note that you can use zoo/xts objects with lm(); 
# you don't *need* a data.frame, but you can't mix types 
# with zoo/xts because they use a matrix internally 
m <- lm(eqn, data=pair) 
beta <- coef(m)[1] 

cat("Assumed hedge ratio is", beta, "\n") 

# index by number, since we won't always know the colnames 
sprd <- pair[,1] - beta*pair[,2] 
ht <- adf.test(sprd, alternative="stationary", k=0) 

cat("ADF p-value is", ht$p.value, "\n") 

if (ht$p.value < 0.05) { 
    cat("The spread is likely mean-reverting\n") 
} else { 
    cat("The spread is not mean-reverting.\n") 
}