CODEWITHSUNDEEP
rxts
khoxsey
khoxsey

Reputation: 1403

Issue with to.weekly parameter drop.time

I am either misunderstanding the docs or having a problem with the drop.time=TRUE parameter in to.weekly(). For a simple example, add a time component onto some sample daily data and roll it up to weekly:

library(xts)
data(sample_matrix)
d <- as.xts(sample_matrix)
index(d) <- index(d)+50

w1 <- to.weekly(d, drop.time=TRUE)
head(w1,1)
                      d.Open   d.High    d.Low  d.Close
2007-01-07 00:00:50 50.03978 50.42188 49.95041 49.99185

w2 <- to.weekly(d, drop.time=FALSE)
head(w2,1)
                      d.Open   d.High    d.Low  d.Close
2007-01-07 00:00:50 50.03978 50.42188 49.95041 49.99185

The docs say:

Setting drop.time to TRUE (the default) will convert a series that includes a time component into one with just a date index, as the time index is often of little value in lower frequency series.

This question mentions that drop.time depends upon indexClass(d)[1] == 'POSIXt' but that appears not to help:

indexClass(d)
[1] "POSIXct" "POSIXt"

indexClass(d) <- c('POSIXt', 'POSIXct')
w3 <- to.weekly(d, drop.time=TRUE, name=NULL)
head(w3,1)
                        Open     High      Low    Close
2007-01-07 00:00:50 50.03978 50.42188 49.95041 49.99185

I'm sure I can just truncate off the time component, but am curious what I'm doing wrong.

Upvotes: 1

Views: 215

Answers (1)

Joshua Ulrich
Joshua Ulrich

Reputation: 176648

This has been fixed in r613 on R-forge. xts:::.drop.time was looking for the virtual POSIXt class in the first position of the class attribute, but (at some point) the order of the POSIXct and POSIXt classes had been switched in base R. I tried to make this check more robust to future changes.

I also made xts:::.drop.time actually change the underlying index. It had been setting indexClass(x) <- "Date", but the underlying index was still the original POSIXt times (including intra-day components). This would cause funky merges.

library(xts)
data(sample_matrix)
d1 <- d2 <- as.xts(sample_matrix)
index(d1) <- index(d1)+50
D1 <- to.daily(d1)
D2 <- to.daily(d2)
head(merge(D1,D2),2)  # old behavior
            d1.Open  d1.High   d1.Low d1.Close  d2.Open  d2.High   d2.Low d2.Close
2007-01-02       NA       NA       NA       NA 50.03978 50.11778 49.95041 50.11778
2007-01-02 50.03978 50.11778 49.95041 50.11778       NA       NA       NA       NA
head(merge(D1,D2),1)  # new behavior
            d1.Open  d1.High   d1.Low d1.Close  d2.Open  d2.High   d2.Low d2.Close
2007-01-02 50.03978 50.11778 49.95041 50.11778 50.03978 50.11778 49.95041 50.11778

Thanks for the report and example!

Upvotes: 1

Related Questions