Exploring NYPD Complaints

NYPD Complaint Data Historic

NYPD complaint trends

library(readr)
library(lubridate)
library(dplyr)
library(ggplot2)

nypd <- read_csv("~/ShinyApps/madlan/nypd.csv")

nypd<-subset(nypd, !is.na(nypd$Latitude))
nypd<-subset(nypd, !is.na(nypd$Longitude))

nypd$year<-year(mdy(nypd$CMPLNT_FR_DT))
nypd<-nypd%>%filter(year>2005)

nypd_trends<-nypd%>%group_by(LAW_CAT_CD, year)%>%summarize(n=sum(!is.na(year)))
ggplot(data=nypd_trends, aes(x=year, y=n)) +
  geom_line(aes(color=LAW_CAT_CD))

nypd_trends_aread<-nypd%>%group_by(BORO_NM,LAW_CAT_CD, year)%>%summarize(n=sum(!is.na(year)))
ggplot(data=nypd_trends_aread, aes(x=year, y=n)) +
  geom_line(aes(color=LAW_CAT_CD))+
  facet_wrap(~BORO_NM)

nypd$hour<-hour(hms(nypd$CMPLNT_FR_TM))
hour_table<-nypd%>%group_by(LAW_CAT_CD, hour)%>%summarize(n=sum(!is.na(hour)))
ggplot(hour_table, aes(x=hour, y=n, colour=LAW_CAT_CD))+geom_line()

hour_table_areas<-nypd%>%group_by(BORO_NM, LAW_CAT_CD, hour)%>%summarize(n=sum(!is.na(hour)))
ggplot(hour_table_areas, aes(x=hour, y=n, colour=LAW_CAT_CD))+geom_line()+facet_wrap(~BORO_NM)

nypd$day <- weekdays(as.Date(nypd$CMPLNT_FR_DT, "%m/%d/%Y"))
day_table<-nypd%>%group_by(LAW_CAT_CD, day)%>%summarize(n=sum(!is.na(day)))
day_table$day <- factor(day_table$day, levels= c("Sunday", "Monday", "Tuesday", "Wednesday", "Thursday", "Friday", "Saturday"))
day_table<-day_table[order(day_table$day),]
ggplot(day_table, aes(x=day, y=n))+ geom_bar(stat = "identity")+facet_wrap(~LAW_CAT_CD)

library(leaflet)
map1<-nypd%>%filter(year==2015, BORO_NM=="MANHATTAN", OFNS_DESC=="ROBBERY")%>%select(Longitude, Latitude, PREM_TYP_DESC)
leaflet(data = map1) %>%
   addTiles() %>%
 addMarkers(clusterOptions = markerClusterOptions(), 
            label=map1$PREM_TYP_DESC)

# box<-make_bbox(data=man_fal, Longitude, Latitude)
# calc_zoom(box)
library(ggmap)
Manhattan_map = qmap("Manhattan", zoom = 13, 
    source="stamen", maptype="toner",darken = c(.3,"#BBBBBB"))
Manhattan_map + geom_point(data=map1, aes(x=Longitude, y=Latitude), 
                              color="dark green", alpha=.15, size=2)

Find effect size:

nypd1<-nypd%>%sample_n(size=1000)
nypd1$day<-factor(nypd1$day, levels = c("Monday", "Tuesday", "Wednesday", "Thursday", "Friday", "Saturday", "Sunday"),
ordered = TRUE)
nypd1$day<-as.integer(nypd1$day)
nypd2<-nypd1%>%select(KY_CD, ADDR_PCT_CD, year, hour, day, X_COORD_CD, Y_COORD_CD)
# mod_1<-lm(KY_CD ~ hour, data = nypd2)
# mod_2<-lm(KY_CD ~ day, data = nypd2)
mod_1<-lm(KY_CD ~ hour+day+ADDR_PCT_CD+year, data = nypd2)
print(mod_1)

Call:
lm(formula = KY_CD ~ hour + day + ADDR_PCT_CD + year, data = nypd2)

Coefficients:
(Intercept)         hour          day  ADDR_PCT_CD         year  
 -3202.2715       0.8427       0.6247      -0.1729       1.7362  

Analysis of Variance Table

anova(mod_1)

Analysis of Variance Table

Response: KY_CD
           Df   Sum Sq Mean Sq F value   Pr(>F)    
hour        1   242659  242659  11.246 0.000828 ***
Residuals 998 21534505   21578                     
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

The analysis above shows a good fitting for the chosen model.