I seem to be getting an error in R when trying to write a simple linear regression based pairs trading code. I suspect this may be an error coming from the downloaded data? However, I'm unsure of whether I am right or how to deal with the bug. As you can probably see, I am relatively new to this sort of thing in R. Any help with this would be extremely appreciated.
symbols <- c("GOLDBEES.NS", "NIFTYBEES.NS")
getSymbols(symbols)
#[1] "GOLDBEES.NS" "NIFTYBEES.NS"
startT <- "2011-01-01"
endT <- "2014-01-01"
rangeT <- paste(startT,"::",endT,sep ="")
tGOLDBEES <- GOLDBEES.NS[,6][rangeT]
tNIFTYBEES <- NIFTYBEES.NS[,6][rangeT]
startO <- "2014-02-01"
endO <- "2016-04-01"
rangeO <- paste(startO,"::",endO,sep ="")
oGOLDBEES <- GOLDBEES.NS[,6][rangeO]
oNIFTYBEES <- NIFTYBEES.NS[,6][rangeO]
pdtGOLDBEES <- diff(tGOLDBEES)[-1]
pdtNIFTYBEES <- diff(tNIFTYBEES)[-1]
model <- lm(pdtGOLDBEES ~ pdtNIFTYBEES - 1)
#Error in model.frame.default(formula = pdtGOLDBEES ~ pdtNIFTYBEES - 1,
#: variable lengths differ (found for 'pdtNIFTYBEES')
As you have already noticed, nrow(GOLDBEES.NS) gives 1742 while nrow(NIFTYBEES.NS) gives 1925. Let's have a closer look:
a <- attr(GOLDBEES.NS, "index")
a <- as.integer((a - a[1]) / 86400) # number of days since 2008-01-01
b <- attr(NIFTYBEES.NS, "index")
b <- as.integer((b - b[1]) / 86400) # number of days since 2008-01-01
You don't really have consecutive daily observations. We should only work with data from common dates.
GOLDBEES.NS <- GOLDBEES.NS[a %in% b]
NIFTYBEES.NS <- NIFTYBEES.NS[b %in% a]
nrow(GOLDBEES.NS) # 1740
nrow(NIFTYBEES.NS) # 1740
Now you can use your code:
startT <- "2011-01-01"
endT <- "2014-01-01"
rangeT <- paste(startT,"::",endT,sep ="")
tGOLDBEES <- GOLDBEES.NS[,6][rangeT]
tNIFTYBEES <- NIFTYBEES.NS[,6][rangeT]
pdtGOLDBEES <- diff(tGOLDBEES)[-1]
pdtNIFTYBEES <- diff(tNIFTYBEES)[-1]
model <- lm(pdtGOLDBEES ~ pdtNIFTYBEES - 1)
#Call:
#lm(formula = pdtGOLDBEES ~ pdtNIFTYBEES - 1)
#Coefficients:
#pdtNIFTYBEES
# -0.6383