20. Dates

Author

Affiliation

Smit, A. J.

University of the Western Cape

This script covers some of the more common issues we may face while dealing with dates.

Dates

# Load libraries
library(tidyverse)
library(lubridate)
library(zoo)

# Load data
sad_dates <- read.csv(here::here("data", "SAMOS", "other_data", "sad_dates.csv"))

1 Date Details

Look at strip time format for guidance

?strptime

Check the local time zone

Sys.timezone(location = TRUE)

R> [1] "Africa/Johannesburg"

2 Creating Daily Dates

Create date columns out of the mangled date data we have loaded.

# Create good date column
new_dates <- sad_dates %>%
  mutate(new_good = as.Date(good))

# Correct bad date column
new_dates <- new_dates %>%
  mutate(new_bad = as.Date(bad, format = "%m/%d/%y"))

# Correct ugly date column
new_dates <- new_dates %>%
  mutate(new_ugly = seq(as.Date("1998-01-13"), as.Date("1998-01-21"), by = "day"))

3 Creating Hourly Dates

If we want to create date values out of data that have hourly values (or smaller), we must create ‘POSIXct’ values because ‘Date’ values may not have a finer temporal resolution than one day.

# Correcting good time stamps with hours
new_dates <- new_dates %>%
  mutate(new_good_hours = as.POSIXct(good_hours, tz = "Africa/Mbabane"))

# Correcting bad time stamps with hours
new_dates <- new_dates %>%
  mutate(new_bad_hours = as.POSIXct(bad_hours, format = "%Y-%m-%d %I:%M:%S %p", tz = "Africa/Mbabane"))

# Correcting bad time stamps with hours
new_dates <- new_dates %>%
  mutate(new_ugly_hours = seq(as.POSIXct("1998-01-13 09:00:00", tz = "Africa/Mbabane"),
                              as.POSIXct("1998-01-13 17:00:00", tz = "Africa/Mbabane"), by = "hour"))

But should not there be a function that loads dates correctly?

4 Importing Dates in One Step

Why yes, yes there is. read_csv() is the way to go.

smart_dates <- read_csv(here::here("data", "SAMOS", "other_data", "sad_dates.csv"))

But why does it matter that we correct the values to dates? For starters, it affects the way our plots look/work. Let us create some random numbers for plotting, see how these compare against our date values when we create figures.

# Generate random number
smart_dates$numbers <- rnorm(9, 2, 10)

# Scatterplot with correct dates
ggplot(smart_dates, aes(x = good, y = numbers)) +
  geom_point() +
  geom_smooth(method = "lm", se = F)

# Scatterplot with incorrect dates
ggplot(smart_dates, aes(x = bad, y = numbers)) +
  geom_point() +
  geom_smooth(method = "lm", se = F)
# OR
ggplot(smart_dates, aes(x = ugly, y = numbers)) +
  geom_point() +
  geom_smooth(method = "lm", se = F)

Figure 1: Dates deeper 1.

Figure 2: Dates deeper 1.

Figure 3: Dates deeper 1.

If the dates are formatted correctly it also allows us to do schnazy things with the data.

smart_dates$good[4]+32

R> [1] "1998-02-17"

smart_dates$good[9]-smart_dates$good[3]

R> Time difference of 6 days

as.Date(smart_dates$good[9]:smart_dates$good[3])

R> [1] "1998-01-21" "1998-01-20" "1998-01-19" "1998-01-18" "1998-01-17"
R> [6] "1998-01-16" "1998-01-15"

smart_dates$good[9]-10247

R> [1] "1970-01-01"