Working with Strings, Dates, and Factors

Dr. Mine Dogucu

Packages

1library(stringr)
2library(lubridate)
3library(forcats)
1
We will use the stringr package for handling strings.
2
We will use the lubridate package for handling dates.
3
We will use the forcats package for handling factors. You can think of forcats as for cat egorical variable s or as an anagram of factors.

Packages

All these packages are part of tidyverse

library(tidyverse)
── Attaching core tidyverse packages ──────────────────────── tidyverse 2.0.0 ──
✔ dplyr   1.2.1     ✔ readr   2.2.0
✔ ggplot2 4.0.2     ✔ tibble  3.3.1
✔ purrr   1.2.2     ✔ tidyr   1.3.2
── Conflicts ────────────────────────────────────────── tidyverse_conflicts() ──
✖ dplyr::filter() masks stats::filter()
✖ dplyr::lag()    masks stats::lag()
ℹ Use the conflicted package (<http://conflicted.r-lib.org/>) to force all conflicts to become errors

Data

sf911 <- 
  read_csv(here::here("data/Fire_Department_and_Emergency_Medical_Services_Dispatched_Calls_for_Service_20260419.csv")) |> 
  janitor::clean_names() |> 
  rename(neighborhoods_boundaries = neighborhooods_analysis_boundaries)

glimpse

glimpse(sf911)
Rows: 369,089
Columns: 36
$ call_number                    <dbl> 250010495, 250010101, 250010606, 250010…
$ unit_id                        <chr> "T06", "T07", "B01", "E03", "M555", "B0…
$ incident_number                <dbl> 25000102, 25000013, 25000123, 25000084,…
$ call_type                      <chr> "Alarms", "Alarms", "Alarms", "Traffic …
$ call_date                      <chr> "01/01/2025", "01/01/2025", "01/01/2025…
$ watch_date                     <chr> "12/31/2024", "12/31/2024", "12/31/2024…
$ received_dt_tm                 <chr> "2025 Jan 01 02:18:24 AM", "2025 Jan 01…
$ entry_dt_tm                    <chr> "2025 Jan 01 02:19:59 AM", "2025 Jan 01…
$ dispatch_dt_tm                 <chr> "2025 Jan 01 02:21:00 AM", "2025 Jan 01…
$ response_dt_tm                 <chr> "2025 Jan 01 02:23:46 AM", "2025 Jan 01…
$ on_scene_dt_tm                 <chr> "2025 Jan 01 02:28:33 AM", "2025 Jan 01…
$ transport_dt_tm                <chr> NA, NA, NA, NA, "2025 Jan 01 01:33:03 A…
$ hospital_dt_tm                 <chr> NA, NA, NA, NA, "2025 Jan 01 01:44:31 A…
$ call_final_disposition         <chr> "Fire", "Fire", "Fire", "Code 2 Transpo…
$ available_dt_tm                <chr> "2025 Jan 01 02:30:48 AM", "2025 Jan 01…
$ address                        <chr> "18TH ST/MARKET ST", "19TH ST/MISSION S…
$ city                           <chr> "San Francisco", "San Francisco", "San …
$ zipcode_of_incident            <dbl> 94114, 94110, 94108, 94102, 94107, 9412…
$ battalion                      <chr> "B05", "B02", "B01", "B02", "B03", "B07…
$ station_area                   <chr> "24", "07", "13", "36", "08", "14", "36…
$ box                            <chr> "5413", "5423", "1313", "1552", "2154",…
$ original_priority              <chr> "3", "3", NA, "A", "3", "3", "3", "3", …
$ priority                       <chr> "3", "3", "3", "3", "3", "3", "3", "3",…
$ final_priority                 <dbl> 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 2, …
$ als_unit                       <lgl> FALSE, FALSE, FALSE, FALSE, TRUE, FALSE…
$ call_type_group                <chr> "Alarm", "Alarm", "Alarm", "Non Life-th…
$ number_of_alarms               <dbl> 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, …
$ unit_type                      <chr> "TRUCK", "TRUCK", "CHIEF", "ENGINE", "M…
$ unit_sequence_in_call_dispatch <dbl> 2, 3, 2, 1, 1, 2, 1, 1, 3, 3, 10, 2, 3,…
$ fire_prevention_district       <chr> "5", "6", "1", "2", "3", "7", "2", "2",…
$ supervisor_district            <dbl> 8, 9, 3, 5, 6, 1, 6, 5, 6, 5, 5, 2, 5, …
$ neighborhoods_boundaries       <chr> "Castro/Upper Market", "Mission", "Chin…
$ row_id                         <chr> "250010495-T06", "250010101-T07", "2500…
$ case_location                  <chr> "POINT (-122.444629428 37.759708374)", …
$ data_as_of                     <chr> "2025 Jan 01 03:32:04 AM", "2025 Jan 01…
$ data_loaded_at                 <chr> "2025 Jan 08 04:18:30 AM", "2025 Jan 08…

Strings

Replace a string

sf911 |> 
  mutate(address_long = str_replace(address, "ST", "STREET")) |>
  select(address_long) |> 
  head(3)
# A tibble: 3 × 1
  address_long          
  <chr>                 
1 18TH STREET/MARKET ST 
2 19TH STREET/MISSION ST
3 CLAY STREET/GRANT AVE

Replace a string

sf911 |> 
  mutate(address_long = str_replace_all(address, "ST", "STREET")) |>
  select(address_long) 
# A tibble: 369,089 × 1
   address_long                                
   <chr>                                       
 1 18TH STREET/MARKET STREET                   
 2 19TH STREET/MISSION STREET                  
 3 CLAY STREET/GRANT AVE                       
 4 FULTON STREET/HYDE STREET/UNITED NATIONS PLZ
 5 02ND STREET/KING STREET                     
 6 31STREET AVE/FULTON STREET                  
 7 09TH STREET/JESSIE STREET                   
 8 POLK STREET/WILLOW STREET                   
 9 06TH STREET/STREETEVENSON STREET            
10 GEARY STREET/JONES STREET                   
# ℹ 369,079 more rows

Replace a string

address_key <- 
  c(
  "ST"   = "STREET", 
  "AVE"  = "AVENUE", 
  "PL"   = "PLACE", 
  "TER"  = "TERRACE"
  )

Replace a string

sf911 |> 
  mutate(address_long = str_replace_all(address, address_key)) |>
  select(address_long)
# A tibble: 369,089 × 1
   address_long                                   
   <chr>                                          
 1 18TH STREET/MARKET STREET                      
 2 19TH STREET/MISSION STREET                     
 3 CLAY STREET/GRANT AVENUE                       
 4 FULTON STREET/HYDE STREET/UNITED NATIONS PLACEZ
 5 02ND STREET/KING STREET                        
 6 31STREET AVENUE/FULTON STREET                  
 7 09TH STREET/JESSIE STREET                      
 8 POLK STREET/WILLOW STREET                      
 9 06TH STREET/STREETEVENSON STREET               
10 GEARY STREET/JONES STREET                      
# ℹ 369,079 more rows

Change letter types

sf911 |> 
  select(address) |> 
  mutate(address = str_to_title(address)) |>
  head()
# A tibble: 6 × 1
  address                             
  <chr>                               
1 18th St/Market St                   
2 19th St/Mission St                  
3 Clay St/Grant Ave                   
4 Fulton St/Hyde St/United Nations Plz
5 02nd St/King St                     
6 31st Ave/Fulton St

We can also use str_to_upper() or str_to_lower() as needed

glue

sf911 |> 
  mutate(
    description = str_glue(
      "Incident {call_number} occurred in {neighborhoods_boundaries}"
      )
    ) |> 
  select(description) |> 
  head(3)
# A tibble: 3 × 1
  description                                       
  <glue>                                            
1 Incident 250010495 occurred in Castro/Upper Market
2 Incident 250010101 occurred in Mission            
3 Incident 250010606 occurred in Chinatown

detect

sf911 |>
  mutate(is_fire_related = str_detect(call_type, "Fire|Smoke")) |>
  select(call_type, is_fire_related) |> 
  slice(705:714)
# A tibble: 10 × 2
   call_type                          is_fire_related
   <chr>                              <lgl>          
 1 Medical Incident                   FALSE          
 2 Medical Incident                   FALSE          
 3 Structure Fire / Smoke in Building TRUE           
 4 Structure Fire / Smoke in Building TRUE           
 5 Medical Incident                   FALSE          
 6 Structure Fire / Smoke in Building TRUE           
 7 Medical Incident                   FALSE          
 8 Medical Incident                   FALSE          
 9 Medical Incident                   FALSE          
10 Medical Incident                   FALSE

Dates

ISO Date

Note

While there are many ways to write a date for humans, there is only one universally accepted format for data as standardized by the International Organization for Standardization (ISO). It follows the order of YYYY-MM-DD (e.g., 1986-12-20). Writing dates this way ensures they are sorted correctly by computers and avoids any international confusion between months and days.

UTC and Timezones

  • UTC stands for Coordinated Universal Time. Think of it as the world’s main clock.

  • UTC is not a timezone but a high-precision time standard that all time zones use to stay synchronized.

  • UTC never observes Daylight Saving Time, it never skips or repeats hours, making it the safest format for storing data and performing math.

  • Irvine is UTC-8 (8 hours behind UTC) between first Sunday of November and second Sunday of March which is specified as Pacific Standard Time (PST) and UTC-7 during rest of the year which is specified as Pacific Daylight Time (PDT).

received_dt_tm

sf911 |> 
  select(received_dt_tm) |> 
  head(3)
# A tibble: 3 × 1
  received_dt_tm         
  <chr>                  
1 2025 Jan 01 02:18:24 AM
2 2025 Jan 01 12:20:36 AM
3 2025 Jan 01 03:00:07 AM

ymd_hms()

sf911 <-
  sf911 |>
  mutate(received_dt_tm = ymd_hms(received_dt_tm))

str(sf911$received_dt_tm)
 POSIXct[1:369089], format: "2025-01-01 02:18:24" "2025-01-01 00:20:36" "2025-01-01 03:00:07" ...

received_dt_tm

sf911 |> 
  select(received_dt_tm) |> 
  head(3)
# A tibble: 3 × 1
  received_dt_tm     
  <dttm>             
1 2025-01-01 02:18:24
2 2025-01-01 00:20:36
3 2025-01-01 03:00:07

with time zone

sf911 <-
  sf911 |> 
  mutate(
    received_dt_tm = with_tz(received_dt_tm, tzone = "America/Los_Angeles")
    )

sf911 |> 
  select(received_dt_tm) |> 
  head(3)
# A tibble: 3 × 1
  received_dt_tm     
  <dttm>             
1 2024-12-31 18:18:24
2 2024-12-31 16:20:36
3 2024-12-31 19:00:07

OlsonNames() function lists different names of timezones.

hour, month and week day

sf911 |> 
  mutate(
    hour_val = hour(received_dt_tm),
    month_val = month(received_dt_tm, label = TRUE),
    day_name = wday(received_dt_tm, label = TRUE)
  ) |> 
  select(received_dt_tm, hour_val, month_val, day_name) |> 
  head(3)
# A tibble: 3 × 4
  received_dt_tm      hour_val month_val day_name
  <dttm>                 <int> <ord>     <ord>   
1 2024-12-31 18:18:24       18 Dec       Tue     
2 2024-12-31 16:20:36       16 Dec       Tue     
3 2024-12-31 19:00:07       19 Dec       Tue

duration

sf911 |>
  mutate(
    goal_arrival_time = received_dt_tm + dseconds(480)
  ) |>
  select(received_dt_tm, goal_arrival_time) |>
  head(3)
# A tibble: 3 × 2
  received_dt_tm      goal_arrival_time  
  <dttm>              <dttm>             
1 2024-12-31 18:18:24 2024-12-31 18:26:24
2 2024-12-31 16:20:36 2024-12-31 16:28:36
3 2024-12-31 19:00:07 2024-12-31 19:08:07

duration

event_start <- ymd_hms("2026-03-07 05:00:00", 
                     tz = "America/Los_Angeles")

event_start + ddays(1)
[1] "2026-03-08 06:00:00 PDT"
event_start + days(1)
[1] "2026-03-08 05:00:00 PDT"

Note

When performing math with dates, lubridate distinguishes between two ways of measuring time: durations and periods. This distinction is critical when your data spans a Daylight Saving Time (DST) transition. The above example shows a hypothetical event start time as March 7th, 2026 at 5:00 in Los Angeles time zone. When you add duration of exactly 1 day using the ddays() function, you are simply adding 86,400 seconds. When 86,400 seconds have passed the clocks would be showing 06:00 am on March 8th, 2026 due to start of Daylight Saving Time. On the other end, the days() function just adds one day to the calendar day without touching the time component.

difftime

sf911 <- 
  sf911 |> 
  mutate(
    on_scene_dt_tm = ymd_hms(on_scene_dt_tm),
    response_time = on_scene_dt_tm - received_dt_tm
  )

str(sf911$response_time)
 'difftime' num [1:369089] 609 364 NA 381 ...
 - attr(*, "units")= chr "secs"

today and now

today()
[1] "2026-04-24"
now()
[1] "2026-04-24 14:21:08 PDT"

Factors

as.factor()

sf911 <-
  sf911 |>
  mutate(unit_type = as.factor(unit_type)) 
  
str(sf911$unit_type)
 Factor w/ 12 levels "AIRPORT","BLS",..: 12 12 3 5 7 3 5 5 5 3 ...

levels()

levels(sf911$unit_type)
 [1] "AIRPORT"        "BLS"            "CHIEF"          "CP"            
 [5] "ENGINE"         "INVESTIGATION"  "MEDIC"          "PRIVATE"       
 [9] "RESCUE CAPTAIN" "RESCUE SQUAD"   "SUPPORT"        "TRUCK"

levels() as seen in ggplot

ggplot(sf911, aes(y = unit_type)) +
  geom_bar()
A horizontal bar chart titled by unit type, showing the count of various emergency service unit types. From bottom to top, the Y-axis lists unit categories including: AIRPORT, BLS, CHIEF, CP, ENGINE, INVESTIGATION, MEDIC, PRIVATE, RESCUE CAPTAIN, RESCUE SQUAD, SUPPORT, and TRUCK. The X-axis represents counts ranging from 0 to 90,000+ (in increments of 30,000). ENGINE appears to have the highest count, exceeding 90,000, while unit types such as AIRPORT, and INVESTIGATION appear to have the lowest counts, close to zero. The chart provides a comparative view of how frequently each unit type appears in the dataset.

Figure 1

fct_infreq()

sf911 <- 
  sf911 |> 
  mutate(unit_type = fct_infreq(unit_type))

str(sf911$unit_type)
 Factor w/ 12 levels "ENGINE","MEDIC",..: 3 3 5 1 2 5 1 1 1 5 ...

levels() as seen in ggplot

ggplot(sf911, aes(y = unit_type)) +
  geom_bar()
A horizontal bar chart showing the count of emergency service unit types, ordered from least to most frequent. The Y-axis lists unit types from lowest (on top) to highest frequency (in the bottom): AIRPORT, INVESTIGATION, RESCUE SQUAD, SUPPORT, BLS, RESCUE, CAPTAIN, CP, CHIEF, PRIVATE, TRUCK, MEDIC, and ENGINE. The X-axis represents counts from 0 to 90,000+ (in increments of 30,000). ENGINE has the highest count, exceeding 90,000, followed by MEDIC and TRUCK. AIRPORT and INVESTIGATION have the lowest counts, close to zero. This frequency-sorted view makes it easy to identify the most and least common unit types in the dataset.

Figure 2

levels() as seen in ggplot with fct_rev()

sf911 |>
  mutate(unit_type = fct_rev(unit_type)) |> 
  ggplot(aes(y = unit_type)) +
  geom_bar()
A horizontal bar chart showing the count of emergency service unit types, ordered from most frequent to least. The Y-axis lists unit types from lowest (in the bottom) to highest frequency (on top): AIRPORT, INVESTIGATION, RESCUE SQUAD, SUPPORT, BLS, RESCUE, CAPTAIN, CP, CHIEF, PRIVATE, TRUCK, MEDIC, and ENGINE. The X-axis represents counts from 0 to 90,000+ (in increments of 30,000). ENGINE has the highest count, exceeding 90,000, followed by MEDIC and TRUCK. AIRPORT and INVESTIGATION have the lowest counts, close to zero. This frequency-sorted view makes it easy to identify the most and least common unit types in the dataset with the most frequent bar on top.

Figure 3

levels() as seen in ggplot with fct_lump_n()

sf911 |>
  mutate(unit_type = fct_lump_n(unit_type, n = 3)) |>
  ggplot(aes(y = unit_type)) +
  geom_bar()
A horizontal bar chart titled unit_type on the y-axis and count on the x-axis. The chart displays the frequency of four categories: Other: The most frequent category, with a count of nearly 120,000. ENGINE: The second most frequent, with a count of approximately 110,000. MEDIC: Slightly lower than ENGINE, with a count of approximately 100,000. TRUCK: The least frequent, with a count of approximately 40,000. The order of the categories go as Other, TRUCK, MEDIC, and ENGINE from top to bottom.

Figure 4

fct_relevel()

sf911 |>
  mutate(unit_type = fct_relevel(unit_type, "ENGINE")) |>
  group_by(unit_type) |>
  summarize(total_als_unit = sum(als_unit))
# A tibble: 12 × 2
   unit_type      total_als_unit
   <fct>                   <int>
 1 ENGINE                  89853
 2 MEDIC                   88065
 3 TRUCK                    2336
 4 PRIVATE                     0
 5 CHIEF                       0
 6 CP                        135
 7 RESCUE CAPTAIN           9821
 8 BLS                         0
 9 SUPPORT                  5059
10 RESCUE SQUAD                0
11 INVESTIGATION               0
12 AIRPORT                     0

fct_relevel()

sf911 |>
  mutate(unit_type = fct_relevel(unit_type, "ENGINE", "TRUCK")) |>
  group_by(unit_type) |>
  summarize(total_als_unit = sum(als_unit))
# A tibble: 12 × 2
   unit_type      total_als_unit
   <fct>                   <int>
 1 ENGINE                  89853
 2 TRUCK                    2336
 3 MEDIC                   88065
 4 PRIVATE                     0
 5 CHIEF                       0
 6 CP                        135
 7 RESCUE CAPTAIN           9821
 8 BLS                         0
 9 SUPPORT                  5059
10 RESCUE SQUAD                0
11 INVESTIGATION               0
12 AIRPORT                     0

fct_reorder()

sf911 <- sf911 |>
  mutate(
    neighborhoods_boundaries = as.factor(neighborhoods_boundaries),
    neighborhoods = fct_reorder(neighborhoods_boundaries, response_time)
    ) 

str(sf911$neighborhoods_boundaries)
str(sf911$neighborhoods)
 Factor w/ 42 levels "Bayview Hunters Point",..: 3 19 4 37 20 27 35 37 35 37 ...
 Factor w/ 42 levels "Nob Hill","Western Addition",..: 15 10 8 12 22 16 24 12 24 12 ...