在 R 中从 Bloomberg 下载多个时间序列的干净方法

Question

i am trying to download some time series data about euro swaps (EUSA10 Currency for example) in R using the blpapi but i am encountering the following problems:

1. if i try to download for example 2y, 5y, 10y and 30y swap rates using the include.non.trading.days=FALSE option, the resulting time series are for some reason of different length and i receive a message error about it. If, on the other hand i set the non trading day option to true i have similar length time series that can then be cleaned up using the na.omit() function

2. the format in which the data is downloaded is messy...i would like to have a data frame in which the first column is the date, second column is the first security, third column is second security and so forth. Instead what i get is [date][security][date][security2]......[date][securityN] . Any suggestions on how to solve this?

Below a quick few lines i wrote as an example

# Load package
library(Rblpapi)
# Connect to Bloomberg
blpConnect()
# Declaring securities
sec<-c("eusa2 curncy", "eusa5 curncy", "eusa10 curncy")
# Declaring field to be dowloaded
flds<-"PX_LAST" 

data<-as.data.frame(bdh(sec,flds,start.date=as.Date("2019-08-18"),end.date=as.Date("2020-08-18"), include.non.trading.days=TRUE"))

Answer 1

It's states in the Rblapi manual that the Rblapi::bdh returns

A list with as a many entries as there are entries in securities; each list contains a data.frame with one row per observations and as many columns as entries in fields. If the list is of length one, it is collapsed into a single data frame. Note that the order of securities returned is determined by the backend and may be different from the order of securities in the securities field.

So I'd suggest you rbind the data then reshape it in order to have the result you want. a fast way to do it is use the data.table::rbindlist function it takes a list as input and returns a data.table containing all entries and if idcol=TRUE then it'll append a .id column showing where the data.frame came from. Also this method will work even if you have different number of rows in the data.frame s resulting from the Rblapi::bdh call.

# Declaring field to be dowloaded
flds<-"PX_LAST" 

# LOADING THE DATA FROM THE API
l <- bdh(sec,flds,start.date=as.Date("2019-08-18"),end.date=as.Date("2020-08-18"), include.non.trading.days=TRUE)

# the names of the securities columns as returned by the api
securities <- paste0("eusa", c(2,5,10,15,30), ".curncy.",flds)

# row binding the resulting list
dt <- data.table::rbindlist(l, idcol=T, use.names=FALSE) 
# idcol=T appends an id column (.id) to the resulting data.table
# use.names=F because the columns of the data.frames are different

# remaking the .id column so it reflects the name of the column that it already had
dt[, .id:= securities[.id] ]


# making a wider data.table
data.table::dcast(dt, eusa2.curncy.date ~ .id, value.var=securities[1]) 
# eusa2.curncy.date is the column that defines a group of observation
# .id the name of the columns
# securities[1] or eusa2.curncy.PX_LAST is the column that contains the values

---

data used

As I don't have access to a bloomberg api endpoint I created this mock data which resemble the output of dbh

col.names <- paste0("eusa", rep(c(2,5,10,15,30),each=2), ".curncy.", rep(c(flds,"date"), 5))
l<-rep(list(data.frame(rnorm(200), 1:200)), 5)

for (i in 1:length(l)) colnames(l[[i]]) <- col.names[(2*i-1):(2*i)]