04-dates.Rmd

```{r include=FALSE}
knitr::opts_chunk$set(collapse=TRUE, comment="#>")
```

# Working with dates and times {#dates-times}




```{r include=FALSE}
library(dplyr)
library(readr)
library(reshape2)
library(tidyr)
library(janitor)
library(lubridate)

# Read in required data using public data.gov extract
road_accidents <- readr::read_rds("data/road_accidents_2017.RDS")

# create character version of date to use in examples
road_accidents$Date1 <- as.character(road_accidents$Date)

```


This chapter provides an overview of working with dates and times, for example extracting year or month from a date, and converting characters to a date.

One of the main packages used to work with dates is **{lubridate}**.

More information can be found on the {lubridate} cheatsheet at the following link: https://www.rstudio.com/resources/cheatsheets/

Date vectors are just vectors of class double with an additional class attribute set as "Date".  

```{r}
DfT_birthday <- lubridate::as_date("1919-08-14")
typeof(DfT_birthday)
attributes(DfT_birthday)
```

If we remove the class using `unclass()` we can reveal the value of the double, which is the number of days since "1970-01-01"^[a special date known as the Unix Epoch], since DfT's birthday is before this date, the double is negative.

```{r}
unclass(DfT_birthday)
```

This chapter will be using the road accident data set:

```{r, echo = FALSE, results='asis'}

knitr::kable(head(dplyr::select(road_accidents, acc_index, Date, Police_Force,  Day_of_Week, Time)),
caption = "Reported Road Accidents, 2017")

```

## Working with dates

### Converting a character to a date 

In R, dates can be converted to a specific date variable type in order to use the variable as a date.

Having a variable as a date means that you can:

 - extract the different elements of the date (year, month etc.)
 - calculate differences between dates

This can be done in the following way:

- Identify the order of the year, month, day and use the appropriate function (ymd, mdy, dmy etc.)

```{r, echo = TRUE}

# convert date to date object

# check class of date
class(road_accidents$Date1)

# look at the date variable and see what order it is in (year-m-d)
# therefore use the ymd function
road_accidents$Date1 <- lubridate::ymd(road_accidents$Date1)

# now check class
class(road_accidents$Date1)

```



### Get year from date

Use the **year** function from {lubridate}:

```{r}

road_accidents$Year <- lubridate::year(road_accidents$Date1)

```

*See Table 5.2 for output*

### Get month from date

Use the **month** function from {lubridate}:

```{r}

road_accidents$Month <- lubridate::month(road_accidents$Date1)

```

*See Table 5.2 for output*

### Get day from date

Use the **day** function from {lubridate}:

```{r}

road_accidents$Day <- lubridate::day(road_accidents$Date1)

```

*See Table 5.2 for output*

### Get weekday from date

Use the **wday** function from {lubridate} to get the weekday label:

```{r}

road_accidents$weekday <- lubridate::wday(road_accidents$Date1)

```

*See Table 5.2 for output*

### Get quarter from date

Use the **quarter** function from {lubridate}:

```{r}

road_accidents$Quarter <- lubridate::quarter(road_accidents$Date1)

```

*See Table 5.2 for output*

```{r, include = FALSE}

dates <- dplyr::select(road_accidents, Date1, Year, Quarter, Month, Day, weekday)

```


```{r, echo = FALSE, results='asis'}

knitr::kable(head(dates),
caption = "Using lubridate to extract time information")

```


### Find difference between two dates


```{r, include = FALSE}

# First create new date column so difference between two dates can be found

road_accidents$Date2 <-round_date(road_accidents$Date1, unit = "month")

```

```{r, echo = FALSE, results='asis'}

knitr::kable(head(dplyr::select(road_accidents, acc_index, Date1, Date2)),
caption = "Find difference between two dates")

```

Use the **as.duration** function to find the duration between two dates. The duration to be measured can be specified:

- dhours
- dweeks
- ddays
- dminutes
- dyears

To find out the number of days difference, the **as.duration** function calculates the duration in seconds so the duration must be divided by the desired duration (ddays) to convert to duration in days.

```{r}

road_accidents$date_diff <- 
lubridate::as.duration(road_accidents$Date2 %--% road_accidents$Date1) / ddays(1)

```

```{r, echo = FALSE, results='asis'}

knitr::kable(head(dplyr::select(road_accidents, acc_index, Date1, Date2, date_diff)),
caption = "Find difference between two dates")

```


The `%--%` operator is used to define an **interval**. So, this code is calculating the duration of the interval between `Date2` and `Date1`.

The number after **ddays** indicates by how many units the duration is (i.e. one day).


### Convert month (integer to character)

{base} R has a useful function which takes the month numbers and converts them to the corresponding text.

```{r}

road_accidents$Month_lab <- month.abb[road_accidents$Month]

```

```{r, echo = FALSE, results='asis'}

knitr::kable(head(dplyr::select(road_accidents, acc_index, Date, Month, Month_lab)),
caption = "Convert month to character")

```

### Convert month (character to integer)

{base} R has a useful function which takes the month text and converts them to the corresponding number.

```{r}
 road_accidents$Month <- match(road_accidents$Month_lab,month.abb)

```


```{r, echo = FALSE, results='asis'}

knitr::kable(head(dplyr::select(road_accidents, acc_index, Date, Month, Month_lab)),
caption = "Convert month character to integer")

```



### Merge separate date information into a date

The {lubridate} package can be used in conjunction with the paste function to combine columns separate date information (e.g. year, month, day) into one date variable.

```{r}

road_accidents$date <- 
paste(road_accidents$Year, road_accidents$Month, road_accidents$Day, sep="-") %>% 
  ymd()%>%
  as.Date()

```

```{r, echo = FALSE, results='asis'}

knitr::kable(head(dplyr::select(road_accidents, acc_index, Date, Year, Month, Day, date)),
caption = "Convert month to character")

```

## Working with date-times

A date-time stores date and time information.


### Converting a character to a date-time

This is similar to converting a character to a date as mentioned above.

This can be done in the following way:

- Identify the order of the year, month, day, and time elements (hour, minute and second or just hour and minute) and use the appropriate function (ymd, mdy, dmy etc.)

```{r, eval = FALSE}

# convert date to date object

# look at the date variable and see what order it is in (year-m-d, hms "2017-11-28 14:00)
# therefore use the ymd_hm
road_accidents$Date_time1 <- lubridate::ymd_hm(road_accidents$Date_time)

```


### Extract date from date time variable

Use the **date** function to extract the date from a date time variable.

The year/month/day information can then be extracted from the date using the code examples above.

```{r, eval = FALSE}


road_accidents$Date2 <- lubridate::date(road_accidents$Date_time)

```

### Convert character to hms (time) variable

Convert time as character into a hms variable so the variable can manipulated as a time object.

This can be done using the **{hms}** package.

```{r}


road_accidents$Time <- hms::as_hms(road_accidents$Time)

```

### Extract hour from time

Use the **hour** function from the {lubridate} package to extract hour information.

```{r}


road_accidents$hour <- lubridate::hour(road_accidents$Time)

```

*See Table 5.8 for output* 

### Extract minute from time

Use the **minute** function from the {lubridate} package to extract minute information.

```{r}


road_accidents$minute <- lubridate::minute(road_accidents$Time)

```

*See Table 5.8 for output* 

### Extract second from time

Use the **second** function from the {lubridate} package to extract second information.

```{r}


road_accidents$second <- lubridate::second(road_accidents$Time)

```

*See Table 5.8 for output* 

```{r, echo = FALSE, results='asis'}

knitr::kable(head(dplyr::select(road_accidents, acc_index, Time, hour, minute, second)),
caption = "Extract time information")

```

### Merge separate time information into one variable

Hour, minute and second variables can be merged to create a time variable, and then converted to hms.

```{r}

# merge seperate time information
road_accidents$time2 <- paste(road_accidents$hour,road_accidents$minute, road_accidents$second, sep=":")

# convert to hms
road_accidents$time3 <- hms::as_hms(road_accidents$time2)

```

```{r, echo = FALSE, results='asis'}

knitr::kable(head(dplyr::select(road_accidents, acc_index, hour, minute, second, time3)),
caption = "Merge time information")

```


### find the difference between two times

Use the {base} r **difftime** function to find the difference between two times.

Note: this can also be used to find the difference in days or weeks.

Also note: the object must be hms/date to be able to calculate the difference.

```{r}

time_first <- hms::as.hms("11:00:00")
time_second <- hms::as.hms("11:05:00")

difference <- difftime(time_first, time_second, "mins" )

```

```{r, echo = TRUE }

difference

```

Change the unit of measurement to get different time differences (for days and weeks you'll need a date rather than a hms).

Units: "secs", "mins", "hours", "days", "weeks"