Reporting on Homelessness: Data Analysis for Journalists

Jour 405v, Jour 5003, Spring 2020

Analysis of San Francisco Police Calls for Service Data

Here is the original dataset: 3,048,797 records

https://data.sfgov.org/Public-Safety/Police-Department-Calls-for-Service/hz9m-tj6z/data

This tutorial uses a subset of this data

The Calls for Service were filtered as follows: CONTAINS homeless, 915, 919, 920: Downloaded 157,237 records 3/31/16 to 11/30/2019. This is 5.1% of all calls in the broader database. File renamed to: SF_311_Jan29.xlsx

Part 1: Quick Start

library(tidyverse)

## ── Attaching packages ─────────────────────────────────────── tidyverse 1.3.0 ──

## ✓ ggplot2 3.2.1     ✓ purrr   0.3.3
## ✓ tibble  2.1.3     ✓ dplyr   0.8.4
## ✓ tidyr   1.0.0     ✓ stringr 1.4.0
## ✓ readr   1.3.1     ✓ forcats 0.4.0

## ── Conflicts ────────────────────────────────────────── tidyverse_conflicts() ──
## x dplyr::filter() masks stats::filter()
## x dplyr::lag()    masks stats::lag()

library(janitor)

## 
## Attaching package: 'janitor'

## The following objects are masked from 'package:stats':
## 
##     chisq.test, fisher.test

library(lubridate)

## 
## Attaching package: 'lubridate'

## The following object is masked from 'package:base':
## 
##     date

Reload Data

SF <- rio::import("https://github.com/profrobwells/HomelessSP2020/blob/master/Data/SF_311_Jan29.xlsx?raw=true", which = "SF Police_Department_Calls_for_")

Clean names, Process dates

SF <- janitor::clean_names(SF)
#Process dates
SF$call_date2 <- ymd(SF$call_date)
SF$year <- year(SF$call_date2)

Process dates

Days <- SF %>% 
  count(call_date2) %>% 
  group_by(call_date2) %>% 
  arrange(desc(n))

Types of Crimes

Types <- SF %>% count(original_crime_type_name) %>% 
  group_by(original_crime_type_name) %>% 
  arrange(desc(n))

Calls by Year

Years <- SF %>% 
  count(year) %>% 
  group_by(year) %>% 
  arrange(desc(year))

Actions Taken

Action <- SF %>% 
  count(disposition) %>% 
  arrange(desc(n))

Part 2: Cleaning & Analysis

Question: How many rows? Columns? Supply a list of the column names

nrow(SF) [1] 157237 > ncol(SF) [1] 14

Process dates, check file types

str(SF)

## 'data.frame':    157237 obs. of  16 variables:
##  $ crime_id                : num  1.9e+08 1.9e+08 1.9e+08 1.9e+08 1.9e+08 ...
##  $ original_crime_type_name: chr  "919" "Homeless Complaint" "Homeless Complaint" "915" ...
##  $ report_date             : POSIXct, format: "2019-01-04" "2019-01-20" ...
##  $ call_date               : POSIXct, format: "2019-01-04" "2019-01-20" ...
##  $ offense_date            : POSIXct, format: "2019-01-04" "2019-01-20" ...
##  $ call_time               : POSIXct, format: "1899-12-31 06:58:00" "1899-12-31 06:19:00" ...
##  $ call_date_time          : POSIXct, format: "2019-01-04 06:58:00" "2019-01-20 06:19:00" ...
##  $ disposition             : chr  "HAN" "HAN" "ADV" "HAN" ...
##  $ address                 : chr  "400 Block Of Jones St" "8th And Market" "Mission St/24th St" "Alabama St/23rd St" ...
##  $ city                    : chr  "San Francisco" NA "San Francisco" "San Francisco" ...
##  $ state                   : chr  "CA" "CA" "CA" "CA" ...
##  $ agency_id               : num  1 1 1 1 1 1 1 1 1 1 ...
##  $ address_type            : chr  "Premise Address" "Geo-Override" "Intersection" "Intersection" ...
##  $ common_location         : chr  NA NA NA NA ...
##  $ call_date2              : Date, format: "2019-01-04" "2019-01-20" ...
##  $ year                    : num  2019 2019 2019 2019 2019 ...

Examine how we have created a new date and year column and how they are formatted differently than the rest We can now perform date and year calculations Create Days Table

Question: Using the summary() function, describe the minimum, maximum, median and mean of calls in the Days table

summary(Days)

##    call_date2               n        
##  Min.   :2016-03-31   Min.   : 10.0  
##  1st Qu.:2017-02-28   1st Qu.: 86.0  
##  Median :2018-01-29   Median :119.0  
##  Mean   :2018-01-29   Mean   :117.3  
##  3rd Qu.:2018-12-30   3rd Qu.:148.0  
##  Max.   :2019-11-30   Max.   :232.0

Between March 31, 2016 and Nov. 30, 2019, San Francisco residents placed an average 117 calls to police complaining about homeless people.

Question: Which day had the most calls? Which day had the least?

Days %>% 
  filter(n == 232)

## # A tibble: 1 x 2
## # Groups:   call_date2 [1]
##   call_date2     n
##   <date>     <int>
## 1 2019-08-15   232

Days %>% 
  filter(n == 10)

## # A tibble: 1 x 2
## # Groups:   call_date2 [1]
##   call_date2     n
##   <date>     <int>
## 1 2016-03-31    10

Examine the types of events

Types <- SF %>% count(original_crime_type_name) %>% 
  group_by(original_crime_type_name) %>% 
  arrange(desc(n))

Question: What are the top five complaints in this data and provide the number of complaints

Types <- SF %>% count(original_crime_type_name) %>% 
  group_by(original_crime_type_name) %>% 
  arrange(desc(n))

Create separate table with just the top five counties’ crime rate: dplyr has a “top_n” function that i find handy

Types <- SF %>% 
  count(original_crime_type_name) %>% 
  top_n(5, n) %>% 
  arrange(desc(n))

Export a table into a spreadsheet (csv is a comma separated file)

write.csv(Days,"Days.csv")

Build a table totalling the number of complaints by year

Years <- SF %>% 
  count(year) %>% 
  group_by(year) %>% 
  arrange(desc(year))

EXERCISE: Grouping by Disposition

Look at the Radio Codes spreadsheet under dispositions

https://data.sfgov.org/api/views/hz9m-tj6z/files/b60ee24c-ae7e-4f0b-a8d5-8f4bd29bf1de?download=true&filename=Radio%20Codes%202016.xlsx

Total by disposition

Action <- SF %>% 
  count(disposition) %>% 
  arrange(desc(n))

Ceate a table with serious infractions described in disposition

Example: Here’s a table filtering the dispositions column to show “no disposition” or “gone on arrival”

Nothing <- SF %>% 
  filter(disposition == "ND" | disposition == "GOA")

Question: Create a table with the serious actions including citations and arrests police took in the dispositions

Arrest, Cited, Criminal Activation, SF Fire Dept Medical Staff engaged

Busted <- SF %>% 
  filter(disposition == "ARR" | disposition == "CIT" | disposition == "CRM" | disposition == "SFD") %>% 
  count(disposition) %>% 
  arrange(desc(n))

EXERCISE - A Basic chart of the crime data

ggplot(Years, aes(x = year, y = n)) + 
  geom_bar(stat = "identity") +
  #coord_flip() +    #this makes it a horizontal bar chart instead of vertical
  labs(title = "Homeless Calls Per Year, San Francisco", 
       subtitle = "SF PD Service Call Data, 3/2016-11/2019",
       caption = "Graphic by Wells",
       y="Number of Calls",
       x="Year")

A chart using a dplyr filtering language

Years %>% 
  filter(year >= 2017) %>% 
  ggplot(aes(x = year, y = n)) +
  geom_bar(stat = "identity") +
  #coord_flip() +    #this makes it a horizontal bar chart instead of vertical
  labs(title = "Homeless Calls After 2017, San Francisco", 
       subtitle = "SF PD Service Call Data, 2017-2019",
       caption = "Graphic by Wells",
       y="Number of Calls",
       x="Year")

A more complex filter

SF %>% 
  filter(!is.na(common_location)) %>% 
  count(common_location) %>% 
  top_n(10, n) %>% 
  ggplot(aes(x = common_location, y = n)) +
  geom_bar(stat = "identity") +
  coord_flip() +    #this makes it a horizontal bar chart instead of vertical
  labs(title = "Popular Spots for Homeless, San Francisco", 
       subtitle = "SF PD Service Call Data, 2016-2019",
       caption = "Graphic by Wells",
       y="Number of Calls",
       x="Places")

Question: Chart the total dispositions.

Filter for at least 100 actions. Add color, export image to Blackboard.

Action %>% 
  filter(n > 100) %>% 
  ggplot(aes(x = reorder(disposition, n), y = n, fill=n)) + 
  geom_bar(stat = "identity", show.legend = FALSE) +
  coord_flip() +    #this makes it a horizontal bar chart instead of vertical
  labs(title = "Action on Homeless Calls, San Francisco", 
       subtitle = "SF PD Service Call Data, 3/2016-11/2019",
       caption = "Graphic by Wells",
       y="Number of Calls",
       x="Action")

Part 3: Cleaning Dispositions

Making our charts less ugly

The disposition column is in cop-speak. We need to clean it up

Step #1: Duplicate the column you want to mess with

SF$disposition1 <- SF$disposition

Rename specific strings. Example:

str_replace_all(test.vector, pattern=fixed(‘-’), replacement=fixed(‘:’) )

Details on string manipulation:

https://dereksonderegger.github.io/570L/13-string-manipulation.html

We can do this to replace ABA with “Abated”

SF$disposition1 <- str_replace_all(SF$disposition1, pattern=fixed('ABA'), replacement=fixed('Abated') )
#Again with ADM
SF$disposition1 <- str_replace_all(SF$disposition1, pattern=fixed('ADM'), replacement=fixed('Admonished') )

We can do that 19 times. OR….

Look at this example using a lookup table to replace all the values
https://stackoverflow.com/questions/50615116/renaming-character-variables-in-a-column-in-data-frame-r

Build a table to translate the Cop Speak to English:

dispo_lkup <- c(ABA="Abated", ADM="Admonish", ADV="Advised", ARR="Arrest", CAN="Cancel", CSA="CPSA", 
                CIT="Cited", CRM="Criminal", GOA="Gone", HAN="Handled", NCR="No_Criminal", ND="No_Dispo", 
                NOM="No_Merit", PAS="PlaceSecure", REP="Report", SFD="Medical", UTL="Unfound", VAS="Vehicle_Secure", '22'="Cancel")

#22="Cancel" was handled differently because it is a numeric value: '22'="Cancel"

#This scans "disposition", finds ABA and replaces with Abated, finds ARR, replaces with Arrest, etc
SF$disposition1 <- as.character(dispo_lkup[SF$disposition])

Rerun Action with disposition1

Action <- SF %>% 
  count(disposition1) %>% 
  arrange(desc(n))

Compare our renamed variables to the original disposition

Action <- SF %>% 
  count(disposition1, disposition) %>% 
  arrange(desc(n))

We have codes not listed on the sheet

NA Not recorded 4339

Get rid of the space

SF$disposition <- gsub("Not recorded", "Not_Recorded", SF$disposition)

Add to the list

dispo_lkup <- c(ABA="Abated", ADM="Admonish", ADV="Advised", ARR="Arrest", CAN="Cancel", CSA="CPSA", 
                CIT="Cited", CRM="Criminal", GOA="Gone", HAN="Handled", NCR="No_Criminal", ND="No_Dispo", 
                NOM="No_Merit", PAS="PlaceSecure", REP="Report", SFD="Medical", UTL="Unfound", 
                VAS="Vehicle_Secure", '22'="Cancel", Not_Recorded="NotRecorded")

Rerun

Action <- SF %>% 
  count(disposition1) %>% 
  arrange(desc(n))

Chart Dispositions

Action %>% 
  filter(n > 100) %>% 
  ggplot(aes(x = reorder(disposition1, n), y = n, fill=n)) + 
  geom_bar(stat = "identity", show.legend = FALSE) +
  coord_flip() +    #this makes it a horizontal bar chart instead of vertical
  labs(title = "Action on Homeless Calls, San Francisco", 
       subtitle = "SF PD Service Call Data, 3/2016-11/2019",
       caption = "Graphic by Wells",
       y="Number of Calls",
       x="Action")

- Parse out police codes from narrative: original_crime_type_name Look at the Types table: some columns have one code, some have two. 919 2879 915 Sleeper 290

Some are separated by a slash 915/919 161

We need to unpack that - Cleaning Sequence

#convert all text to lowercase
SF$crime1 <- tolower(SF$original_crime_type_name)
#Replace / with a space
SF$crime1 <- gsub("/", " ", SF$crime1)
#Replace '
SF$crime1 <- gsub("'", "", SF$crime1)
#fix space in homeless complaint
SF$crime1 <- gsub("homeless complaint", "homeless_complaint", SF$crime1)
#split data into two columns
SF <- separate(data = SF, col = crime1, into = c("crime2", "crime3", "crime4"), sep = " ", extra = "merge", fill = "right")

Look at the categories now

Types2 <- SF %>% count(crime2) %>% 
  group_by(crime2) %>% 
  arrange(desc(n))

Question Take the top 10 crime categories from Type2
Relabel them from the numeric radio codes into English
Using the technique earlier in “Build a table to translate the Cop Speak to English”
Relabel the offenses

clean <- c(homeless_complaint="homeless_complaint", '915'="homeless_call", '919'="sit_lying", '920'="aggress_solicit", '915s'="homeless_call", '915x'="homeless_call", drugs="drugs", '601'="trespasser",
           poss="poss", aggressive="aggressive", '811'="intoxicated")

SF$crime2 <- as.character(clean[SF$crime2])

Look at the categories now

Types2 <- SF %>% count(crime2) %>% 
  group_by(crime2) %>% 
  arrange(desc(n))

Question: Make a chart from your cleaned data

Basic chart but with a messed up x axis

Types2 %>% 
  ggplot(aes(x = crime2, y = n, fill=n)) + 
  geom_bar(stat = "identity") +
  coord_flip() +    #this makes it a horizontal bar chart instead of vertical
  labs(title = "Top 10 Homeless Complaints, San Francisco", 
       subtitle = "SF PD Service Call Data, 3/2016-11/2019",
       caption = "Graphic by Wells",
       y="Number of Calls",
       x="Complaint")

Chart with a fixed x axis scale; No values filtered out; Labels added to bars

Types2 %>% 
  filter(!is.na(crime2)) %>% 
  #filter(crime2!=" ") %>%  - a crude alternative to previous line!
  ggplot(aes(x = reorder(crime2, n), y = n, fill=n)) + #reorder sorts the bars
  geom_bar(stat = "identity", show.legend = FALSE) +
  geom_text(aes(label = n), hjust = -.1, size = 3) +
  scale_y_continuous(limits=c(0, 175000)) + #fixes scientific notation
  coord_flip() +    #this makes it a horizontal bar chart instead of vertical
  labs(title = "Top 10 Homeless Complaints, San Francisco", 
       subtitle = "SF PD Service Call Data, 3/2016-11/2019",
       caption = "Graphic by Wells",
       y="Number of Calls",
       x="Complaint")

Part 4: Using Mutate, Pct Calcs

mutate - Create new column(s) in the data, or change existing column(s).

mutate() adds new variables and preserves existing

Example: mtcars <- as.data.frame(mtcars) View(mtcars)

mtcars2 <- mtcars %>% as_tibble() %>% mutate( cyl2 = cyl * 2, cyl4 = cyl2 * 2 )

Process dates using lubidate

SF <- SF %>% 
  mutate(yearmo = format(call_date, "%Y-%m"))

Chart the number of calls by year and month

SF %>% 
  count(yearmo) %>% 
  group_by(yearmo) %>% 
  ggplot(aes(x = yearmo, y = n, fill=n)) +
  geom_bar(stat = "identity") +
  theme(axis.text.x = element_text(angle=90)) +
  #Changes angle of x axis labels
  #coord_flip() +    #this makes it a horizontal bar chart instead of vertical
  labs(title = "Homeless Calls After 2017, San Francisco", 
       subtitle = "SF PD Service Call Data by Month 2017-2019",
       caption = "Graphic by Wells",
       y="Number of Calls",
       x="Year")

Percentage change per month

PCT_CHG_CALLS <- SF %>% 
  select(original_crime_type_name, disposition, address, call_date2, yearmo) %>% 
  count(yearmo) %>% 
  mutate(difference = (n-lag(n))) %>% 
  mutate(pct_change = (difference/abs(lag(n)))*100)

Use grepl to search and tabulate

grep and grepl: see ??grep

http://www.endmemo.com/program/R/grepl.php

Cleaning Sequence

#convert all text to lowercase
SF$crime1 <- tolower(SF$original_crime_type_name)

Search for term, rename, put in new column called “cleaned”

x915 <- SF %>% 
  filter(grepl ("915", original_crime_type_name)) %>% 
  mutate(cleaned = "homeless_complaint")

x919 <- SF %>% 
  filter(grepl ("919", original_crime_type_name)) %>% 
  mutate(cleaned = "sitting_lying")

xsleep <- SF %>% 
  filter(grepl ("sleep", original_crime_type_name)) %>% 
  mutate(cleaned = "sleep")

xaggr <- SF %>% 
  filter(grepl ("aggr", original_crime_type_name)) %>% 
  mutate(cleaned = "aggressive")

xdrug <- SF %>% 
  filter(grepl ("drug", original_crime_type_name)) %>% 
  mutate(cleaned = "drug")

xhomeless <- SF %>% 
  filter(grepl ("homeless_complaint", crime2)) %>% 
  mutate(cleaned = "homeless_complaint")

#Moe, Brooke's Work: 
xnoise <- SF %>% 
  filter(grepl ("415", original_crime_type_name)) %>% 
  mutate(cleaned = "noise")

xposs <- SF %>% 
  filter(grepl ("poss", original_crime_type_name)) %>% 
  mutate(cleaned = "possession")

xtrespasser <- SF %>% 
  filter(grepl ("601", original_crime_type_name)) %>% 
  mutate(cleaned = "trespasser")

xsolicit <- SF %>% 
  filter(grepl ("920", original_crime_type_name)) %>% 
  mutate(cleaned = "solicit")

xinterview <- SF %>% 
  filter(grepl ("909", original_crime_type_name)) %>% 
  mutate(cleaned = "interview")

xtent <- SF %>%
  filter(grepl ("tent", crime1)) %>%
  mutate(cleaned="tent")

xdog <- SF %>%
  filter(grepl ("dog", crime1)) %>%
  mutate(cleaned="dog")

xchopshop <- SF %>%
  filter(grepl ("chop shop", crime1)) %>%
  mutate(cleaned="chopshop")

xpanhandling <- SF %>%
  filter(grepl ("panhandling", crime1)) %>%
  mutate(cleaned="panhandling")

xmusic <- SF %>%
  filter(grepl ("music", crime1)) %>%
  mutate(cleaned="music")

Create new dataframe using rbind

new_total <- rbind(xhomeless, x915, x919, xaggr, xdrug, xsleep, xnoise, xposs, xtrespasser, xsolicit, xinterview, xtent, xdog,
                   xchopshop, xpanhandling, xmusic)

Count it up!

Total_Calls_Master <- new_total %>% 
  count(cleaned) %>% 
  arrange(desc(n))
#rename columns
colnames(Total_Calls_Master)[1:2] <- c("Complaints", "Number")
#export
write_csv(Total_Calls_Master, "Total_Calls_Master.csv")

Make into html table

#install.packages("kableExtra")
library(kableExtra)

## 
## Attaching package: 'kableExtra'

## The following object is masked from 'package:dplyr':
## 
##     group_rows

#This makes html tables called "kables"
Total_Calls_Master %>% 
  kable() %>%
  kable_styling("striped")

Complaints	Number
homeless_complaint	153895
sitting_lying	3282
solicit	262
trespasser	100
sleep	77
interview	75
drug	52
noise	27
aggressive	18
dog	12
tent	9
music	8
chopshop	7
panhandling	5
possession	5

Export from Viewer as .png

Task: Tabulate complaints by day of the week

https://github.com/profrobwells/Data-Analysis-Class-Jour-405v-5003/blob/master/Readings/dealing-with-dates.pdf

SF <- SF %>% 
  mutate(weekday = wday(call_date, label=TRUE, abbr=FALSE))

Build a summary table with the days of the week with the greatest number of calls. Create a graphic. Then build a table to see if the complaints vary by day

Below from Matthew Moore, Katy Seiter, Wells edited

SF <- SF %>% 
  mutate(weekday = wday(call_date, label=TRUE, abbr=FALSE))

Weekday_Count <- SF %>%
  select(weekday, crime_id) %>%
  count(weekday) %>%
  arrange(desc(n))

Graphic of calls by weekdays

Weekday_Count %>% 
  ggplot(aes(x = weekday, y = n, fill=n)) +
  geom_bar(stat = "identity", show.legend = FALSE) +
  theme(axis.text.x = element_text(angle=90)) +
  #Changes angle of x axis labels
  #coord_flip() +    #this makes it a horizontal bar chart instead of vertical
  labs(title = "Homeless Calls By Weekday in San Francisco", 
       subtitle = "SF PD Service Call Data 2017-2019",
       caption = "Graphic by Moore and Seiter",
       y="Number of Calls",
       x="Weekday")

Create a Bubble graphic

ggplot(data = Weekday_Count) + 
  geom_point(mapping = aes(x = weekday, y = n, size = n, color = n), show.legend = FALSE) +
  theme(axis.text.x = element_text(angle=90)) +
  labs(title = "Homeless By Weekday in San Francisco", 
       subtitle = "SF PD Service Call Data 2017-2019: Source: SFPD",
       caption = "Graphic by Moore and Seiter",
       y="Number of Calls",
       x="Weekday")

Improved bubble chart

ggplot(Weekday_Count, aes(x = weekday, y = n)) +
  xlab("Weekday") +
  ylab("Number of Calls") +
  theme_minimal(base_size = 12, base_family = "Georgia") +
  geom_point(aes(size = n, color = n), alpha = 0.7, show.legend = FALSE) +
  scale_size_area(guide = FALSE, max_size = 15) +
  labs(title = "Homeless By Weekday in San Francisco", 
       subtitle = "SF PD Service Call Data 2017-2019: Source: SFPD",
       caption = "Graphic by Moore and Seiter")

- Task #3: Calls vs Dispositions

What calls resulted in arrests? What calls resulted in citations?

Action2 <- SF %>%
  select(crime_id, original_crime_type_name, disposition)

We need to pair the crime type and disposition and then count them

From Michael Adkison:

callsarrest <- Action2 %>% 
  filter(grepl("ARR", disposition)) %>% 
  mutate(cleaned = "Arrest")

To quickly format into percents, load formattable

#install.packages("formattable")
library(formattable)

callsarrest2 <- callsarrest %>% 
  arrange(original_crime_type_name, disposition) %>% 
  count(original_crime_type_name) %>% 
#mutate(PctTotal = (n/441)) %>% 
  arrange(desc(n))

colnames(callsarrest2)[1:2] <- c("Complaints", "Arrests")

Build a table to translate the Cop Speak to English:

clean <- c('Homeless Complaint'="homeless_complaint", homeless_complaint="homeless_complaint", '915'="homeless_complaint", 
           '919'="Sit_lying", '920'="Aggress_solicit", '915s'="homeless_complaint", '915x'="homeless_complaint", 
           drugs="drugs", '601'="trespasser", poss="poss", aggressive="aggressive", '811'="intoxicated", 
           'Drugs / 915'="Drugs", 'Drugs/915'="Drugs")

This scans “disposition”, finds ABA and replaces with Abated, finds ARR, replaces with Arrest, etc callsarrest2\(Complaints <- as.character(clean[callsarrest2\)Complaints])

callsarrest3 <- callsarrest2 %>% 
  select(Complaints, Arrests) %>% 
  group_by(Complaints) %>% 
  summarise(total = sum(Arrests)) %>% 
  mutate(PctTotal = (total/441)) %>% 
  arrange(desc(total))

colnames(callsarrest3)[2] <- "Arrests" 
callsarrest3$PctTotal <- percent(callsarrest3$PctTotal)

#This makes kables
callsarrest3 %>% 
  kable() %>%
  kable_styling("striped")

Complaints	Arrests	PctTotal
Homeless Complaint	412	93.42%
915	19	4.31%
919	4	0.91%
920	2	0.45%
601 / 915	1	0.23%
915 / 909	1	0.23%
Drugs / 915	1	0.23%
Drugs/915	1	0.23%

Part 5: Trends over time

Question: What were the common days for arrests?

SF %>%
  select(weekday, crime_id, disposition) %>%
  filter(grepl("ARR", disposition)) %>%
  count(weekday)

## # A tibble: 7 x 2
##   weekday       n
##   <ord>     <int>
## 1 Sunday       47
## 2 Monday       63
## 3 Tuesday      79
## 4 Wednesday    77
## 5 Thursday     74
## 6 Friday       56
## 7 Saturday     45

Make bubble chart

SF %>%
  select(weekday, crime_id, disposition) %>%
  filter(grepl("ARR", disposition)) %>%
  count(weekday) %>% 
  ggplot(aes(x = weekday, y = n)) +
  xlab("Weekday") +
  ylab("Arrests") +
  theme_minimal(base_size = 12, base_family = "Georgia") +
  geom_point(aes(size = n, color = n), alpha = 0.7, show.legend = FALSE) +
  scale_size_area(guide = FALSE, max_size = 15) +
  labs(title = "Homeless Arrests By Weekday in San Francisco", 
       subtitle = "SF PD Service Call Data 2017-2019: Source: SFPD",
       caption = "Graphic by Wells")

Question: What is the trend for arrests over the time period?

SF %>% 
  filter(grepl("ARR", disposition)) %>%
  count(yearmo) %>% 
  group_by(yearmo) %>% 
  ggplot(aes(x = yearmo, y = n, fill=n)) +
  geom_bar(stat = "identity", show.legend = FALSE) +
  geom_smooth(method = lm, se=FALSE, color = "red") +
  theme(axis.text.x = element_text(angle=90)) +
  #Changes angle of x axis labels
  #coord_flip() +    #this makes it a horizontal bar chart instead of vertical
  labs(title = "Arrest Trends on Homeless Calls in San Francisco", 
       subtitle = "Arrests Based on SF PD Service Call Data by Month 2017-2019",
       caption = "Graphic by Wells",
       y="Number of Calls",
       x="Year")

Question: What are the hours most likely for complaints?

#format to hours
SF$hour <- hour(SF$call_date_time)

SF %>% 
  count(hour) %>% 
  group_by(hour) %>% 
  ggplot(aes(x = hour, y = n, fill=n)) +
  geom_bar(stat = "identity", show.legend = FALSE) +
  theme(axis.text.x = element_text(angle=90)) +
  #Changes angle of x axis labels
  #coord_flip() +    #this makes it a horizontal bar chart instead of vertical
  labs(title = "Hours of Homeless Calls, San Francisco", 
       subtitle = "SF PD Service Call Data by Month 2017-2019",
       caption = "Graphic by Wells",
       y="Number of Calls",
       x="Hour")

- Question: Examine some of the charting options on this tutorial and adapt them to this data using any chart you want # https://paldhous.github.io/wcsj/2017/

SFPD Calls for Service_Feb 23,2020

Rob Wells

2/23/2020