Lagging 1 year in xts object with weekly frequency

Question

I am trying to lag an xts dataset by 1 year. The data has weekly frequency, and I am currently using lag.xts function like so:

lag.xts(my_data, 52)

The problem is not all years have 52 weeks (some have 53). Is there a way I can take this into account somehow?

Answer 1

I'm assuming you want to find the date which is closest to the specified date, minus a year?

This does the equivalent, but for a dataframe of values:

library(tidyverse)

data <- tibble(x = rnorm(100), 
               dates = seq(as.Date("2015-12-27"), length = 100, by = "weeks"))

find_closest_day <- function(t) {
    ty = t - years(1)
    t52 = abs((t - weeks(52)) - ty)
    t53 = abs((t - weeks(53)) - ty)
    if (t52 < t53) {
        return(t - weeks(52))
    } else {
        return(t - weeks(53))
    }
}
find_closest_day<- Vectorize(find_closest_day)

data %>% 
  mutate(lagged_monday = as.Date(find_closest_day(dates))) %>%
  left_join(data, by = c("lagged_monday" = "dates"))

       x.x dates      lagged_monday   x.y
     <dbl> <date>     <date>        <dbl>
 1 -0.0768 2015-12-27 2014-12-28       NA
 2 -0.711  2016-01-03 2015-01-04       NA
 3 -1.94   2016-01-10 2015-01-11       NA
 4  0.584  2016-01-17 2015-01-18       NA
 5 -0.266  2016-01-24 2015-01-25       NA
 6 -1.04   2016-01-31 2015-02-01       NA
 7 -0.954  2016-02-07 2015-02-08       NA
 8 -0.809  2016-02-14 2015-02-15       NA
 9  0.458  2016-02-21 2015-02-22       NA
10  0.367  2016-02-28 2015-03-01       NA
# ℹ 90 more rows