Skip to content

Latest commit

 

History

History
428 lines (324 loc) · 9.2 KB

cm109-data-time-dates.md

File metadata and controls

428 lines (324 loc) · 9.2 KB

time and dates


Aztec/Mayan Calendar by Kim Alaniz is licensed under CC BY 2.0

contents

introduction
prerequisites
the Date class
the POSIXct class
creating date-time objects
decimal dates
math with date-times
references

introduction

Dealing with time and dates is important because so much data is collected with date or time variables, for example, economic indicators, climate data, medical data, etc.

However, dates and times are hard “because they have to reconcile two physical phenomena (the rotation of the Earth and its orbit around the sun) with a whole raft of geopolitical phenomena including months, time zones, and Daylight Savings Time” (Wickham and Grolemund, 2017, Ch. 16). Adding to these inherent difficulties, data are often messing, intervals may be irregular, and different organizations use different formats, e.g., “2019-04-16”, “16 Apr 2019”, “04/16/2019”

To add to our confusion, R includes more than one class of time series variables. We will generally use class Date, but we will sometimes encounter class POSIXct (POSIX calendar time) and class ts (time series). Time-Series objects are not dates precisely, so they are discussed in a separate tutorial.

  • class Date are calendar dates as the number of days since the beginning of 1970
  • class POSIXct are date+time as the number of seconds since the beginning of 1970
  • class ts are time-series objects, typically numeric data vectors, that are observed at equally-spaced time intervals in time.
  • class numeric for decimal dates


▲ top of page

prerequisites

Project setup

  • Start every work session by launching the RStudio Project file for the course, e.g., portfolio.Rproj
  • Ensure your project directory structure satisfies the course requirements

Ensure you have installed the following packages. See install packages for instructions if needed.

  • tidyverse
  • lubridate
library("tidyverse")
library("lubridate")


▲ top of page

the Date class

Like factors, Date objects print as character strings but are stored as integers.

In this example, I start with a character vector,

x <- c("2019-03-01", "2019-06-01", "2019-09-01", "2019-12-01")
typeof(x)
#> [1] "character"

Convert it to a Date class using lubridate::ymd(),

x_date <- ymd(x)
attributes(x_date)
#> $class
#> [1] "Date"

And though the variable appears to be a character vector,

x_date
#> [1] "2019-03-01" "2019-06-01" "2019-09-01" "2019-12-01"

Dates are encoded numerically,

typeof(x_date)
#> [1] "double"

If we unclass() the variable, we reveal the hidden numbers. The Date class is internally stored as the number of days since 1970-01-01,

unclass(x_date)
#> [1] 17956 18048 18140 18231

To confirm that 1970-01-01 is the zero date, try this

x_date <- ymd("1969-12-01", "1970-01-01", "1970-02-01")

unclass(x_date)
#> [1] -31   0  31

To learn more about this class, run ? Dates.


▲ top of page

the POSIXct class

The POSIXct class stores dates and times as the number of seconds since 1970-01-01 00:00:00 GMT. Like factors, POSIXct objects print as character strings but are stored as numbers.

The base R function as.POSIXct() converts date-time strings to POSIXct class.

x_date <- as.POSIXct("2019-04-17 11:24:09 EDT")

The class is POSIXct

attributes(x_date)
#> $class
#> [1] "POSIXct" "POSIXt" 
#> 
#> $tzone
#> [1] ""

When printed, the vector appears as character strings of both calendar date and time in hh:mm:ss format.

x_date
#> [1] "2019-04-17 11:24:09 EDT"

The dates are encoded numerically

typeof(x_date)
#> [1] "double"

If we unclass() the variable, we reveal the hidden numbers. The POSIXct class is internally stored as the number of seconds since 1970-01-01,

unclass(x_date)
#> [1] 1555514649
#> attr(,"tzone")
#> [1] ""

To learn more about this date-time class and the many functions available, run ? DateTimeClasses.

creating date-time objects

Lubridate has some simple functions for converting the most common date strings into Date objects.

  • ymd()
# string
(x <- "2017-01-31")
#> [1] "2017-01-31"
class(x)
#> [1] "character"

# date
(z <- ymd(x))
#> [1] "2017-01-31"
class(z)
#> [1] "Date"
unclass(z)
#> [1] 17197
  • mdy()
# character
(x <- "January 31st, 2017")
#> [1] "January 31st, 2017"

# date using mdy()
(z <- mdy(x))
#> [1] "2017-01-31"
class(z)
#> [1] "Date"

# character
(x <- "01/31/2017")
#> [1] "01/31/2017"

# date using mdy()
(z <- mdy(x))
#> [1] "2017-01-31"
class(z)
#> [1] "Date"
  • dmy()
# character 
(x <- "31-Jan-2017")
#> [1] "31-Jan-2017"

# date using dmy()
(z <- dmy(x))
#> [1] "2017-01-31"
class(z)
#> [1] "Date"
  • make_date() when year, month, and day are separate variables
# character
yyyy <- "2017" 
mm   <- "1"
dd   <- "31" 

# date using make_date() 
(z <- make_date(year = yyyy, month = mm, day = dd))
#> [1] "2017-01-31"
class(z)
#> [1] "Date"

# numeric
yyyy <- 2017 
mm   <- 1
dd   <- 31 

# date using make_date() 
(z <- make_date(year = yyyy, month = mm, day = dd))
#> [1] "2017-01-31"
class(z)
#> [1] "Date"
  • character date and time produce a POSIXct object,
# character in mdyhm form
(x <- c("May 11, 1996 12:05", "September 12, 2001 1:00", "July 1, 1988 3:32"))
#> [1] "May 11, 1996 12:05"      "September 12, 2001 1:00"
#> [3] "July 1, 1988 3:32"

# POSIXct using mdy_hm() 
(x_posix <- mdy_hm(x))
#> [1] "1996-05-11 12:05:00 UTC" "2001-09-12 01:00:00 UTC"
#> [3] "1988-07-01 03:32:00 UTC"
class(x_posix)
#> [1] "POSIXct" "POSIXt"

# character in ymdhms
(x <- c("2019-04-17 3:32:15"))
#> [1] "2019-04-17 3:32:15"

# POSIXct using mdy_hm() 
(x_posix <- ymd_hms(x))
#> [1] "2019-04-17 03:32:15 UTC"
class(x_posix)
#> [1] "POSIXct" "POSIXt"
  • as_date() converts a POSIXct object to a Date object, keeping the calendar date and losing the time information.
# POSIXct
x_posix
#> [1] "2019-04-17 03:32:15 UTC"
class(x_posix)
#> [1] "POSIXct" "POSIXt"
unclass(x_posix)
#> [1] 1555471935
#> attr(,"tzone")
#> [1] "UTC"

# Date using as_date()  
(x_date <- as_date(x_posix))
#> [1] "2019-04-17"
class(x_date)
#> [1] "Date"
unclass(x_date)
#> [1] 18003

For more on creating dates, read 16.2 Creating date/times in the text.


▲ top of page

decimal dates

Decimal dates are floating point numbers yyyy.nnn where nnn represents the time, day, and month to the nearest 1/1000 of the year. For example, 9am on 2019-04-18 to the nearest 1/1000 year is 2019.294.

You can create decimal dates from POSIXct date using decimal_date().

(x <- "2019-04-18 00:00 EDT")
#> [1] "2019-04-18 00:00 EDT"

(x_posix   <- ymd_hm(x, tz = "US/Eastern"))
#> [1] "2019-04-18 EDT"

(x_decimal <- decimal_date(x_posix))
#> [1] 2019.293
class(x_decimal)
#> [1] "numeric"

Decimal dates have a finite resolution. For example, if I add 4 hours to the previous date, I get the same decimal result. But adding 5 hours, and I get the next decimal. So the uncertainty in a decimal date is between 4 and 5 hours.

# the previous date-time plus 3 hours
x <- "2019-04-18 04:00 EDT"
x_posix   <- ymd_hm(x, tz = "US/Eastern")
(x_decimal <- decimal_date(x_posix))
#> [1] 2019.293

# add one more hour
x <- "2019-04-18 05:00 EDT"
x_posix   <- ymd_hm(x, tz = "US/Eastern")
(x_decimal <- decimal_date(x_posix))
#> [1] 2019.294

If you encounter decimal dates in a data frame, they can be converted to POSIXct using date_decimal() and then to Date class using as _date().

(x_decimal <- 219.294) 
#> [1] 219.294

(x_posix <- date_decimal(x_decimal))
#> [1] "0219-04-18 07:26:24 UTC"
class(x_posix)
#> [1] "POSIXct" "POSIXt"

(x_date <- as_date(x_posix))
#> [1] "0219-04-18"
class(x_date)
#> [1] "Date"

math with date-times

The best place to get started doing math with date-time objects is the lubridate cheat sheet. The essential operations are computing

  • periods
  • durations
  • intervals

references

Wickham H and Grolemund G (2017) R for Data Science. O’Reilly Media, Inc., Sebastopol, CA https://r4ds.had.co.nz/

▲ top of page
◁ calendar
◁ index