In business-science/tibbletime: Time Aware Tibbles

Introducing as_period()

Often with time series you want to aggregate your dataset to a less granular period. An example of this might be moving from a daily series to a monthly series to look at broader trends in your data. as_period() allows you to do exactly this.

The period argument in as_period() for specifying the transformation you want is a character with a general format of "frequency period" where frequency is a number like 1 or 2, and period is an interval like weekly or yearly. There must be a space between the two.

Datasets required

library(tibbletime)
library(dplyr)

# Facebook stock prices.
data(FB)

# Convert FB to tbl_time
FB <- as_tbl_time(FB, index = date)

# FANG stock prices
data(FANG)

# Convert FANG to tbl_time and group
FANG <- as_tbl_time(FANG, index = date) %>%
  group_by(symbol)

Daily to monthly

To see this in action, transform the daily FB data set to monthly data.

as_period(FB, '1 month')

# Additionally, the following are equivalent
# as_period(FB, 'month')
# as_period(FB, 'm')
# as_period(FB, '1 m')

Generic periods

You aren't restricted to only 1 month periods. Maybe you wanted every 2 months?

as_period(FB, '2 m')

Or maybe every 25 days? Note that the dates do not line up exactly with a difference of 25 days. This is due to the data set not being completely regular (there are gaps due to weekends and holidays). as_period() chooses the first date it can find in the period specified.

as_period(FB, '25 d')

Details and the `start_date` argument

By default, the date that starts the first group is calculated as:

1) Find the minimum date in your dataset.

2) Floor that date to the period that you specified.

In the 1 month example above, 2013-01-02 is the first date in the series, and because "month" was chosen, the first group is defined as (2013-01-01 to 2013-01-31).

Occasionally this is not what you want. Consider what would happen if you changed the period to "every 2 days". The first date is 2013-01-02, but because "day" is chosen, this isn't floored to 2013-01-01 so the groups are (2013-01-02, 2013-01-03), (2013-01-04, 2013-01-05) and so on. If you wanted the first group to be (2013-01-01, 2013-01-02), you can use the start_date argument.

# Without start_date
as_period(FB, '2 d')

# With start_date
as_period(FB, '2 d', start_date = "2013-01-01")

The `side` argument

By default, the first date per period is returned. If you want the end of each period instead, specify the side = "end" argument.

as_period(FB, 'y', side = "end")

Grouped datasets

One of the neat things about working in the tidyverse is that these functions can also work with grouped datasets. Here we transform the daily series of the 4 FANG stocks to a periodicity of every 2 years.