Between the years 2015 - 2020, was there a upward or
downward trend in accidents involving alcohol?
An analysis using data provided by the Victorian
State Government
Wong Yi Wei s3966890
Last updated: 16 October, 2022
Introduction
There are many accidents happening everyday globally, and factors
such as driving under the influence such as alcohol could increase the
amount of accidents happening.
With an increased governance and more roadblocks happening from the
years since 2015, we will be exploring and analyzing on whether or not
there is a upward or downward trend in fatal accidents involving alcohol
consumptionIntroduction Cont.
This analysis will be using the data set: “Road Crashes for five
Years - Victoria”
Problem Statement
- This presentation aims to explore and analyze on whether there is a
upward or downward trend in fatal accidents involving alcohol
consumption for the years 2010 - 2015.
- We will be using hypothesis testing, as well as statistical graphs
such to make an decision on our conclusion.
Data Cont.
In this data, there are 75320 observations and 65
variables.
The important variables that we will be using are listed as
below:
ACCIDENT_DATE
- This would show us the year on when the accident had happened.
SEVERITY
- This variable shows whether or not the accident was fatal.
FATALITY
- This variable accounts for the fatality and lives lost in the
accident.
ALCOHOL_RELATED
- This variable tells us whether if the accident was caused by alcohol
consumption.
Data Cont.
This data set was imported into R Studio using the read.csv()
function, and the following code head() shows the first 6 of the
data.
crashes<-read.csv(file="C:/Users/ethan/OneDrive - RMIT University/applied analytics assignment 2/Road_Crashes_for_five_Years_Victoria.csv")
head(crashes)
Data Cont.
To preprocess the data and get it ready for analysis and
visualisation, we will first check the data structure using the str()
function. Then, we will make any changes necessary for the data to solve
our problem statement. From the output below, we can see that there are
75320 observations with 65 variables and we need to filter and subset
the variables ACCIDENT_DATE, SEVERITY, FATALITY, and
ALCOHOL_RELATED.
## 'data.frame': 75320 obs. of 65 variables:
## $ X : num 2520841 2526250 2538721 2516548 2485113 ...
## $ Y : num 2392194 2381737 2400595 2389749 2413935 ...
## $ OBJECTID : int 1 2 3 4 5 6 7 8 9 10 ...
## $ ACCIDENT_NO : chr "T20150013821" "T20150013832" "T20150013882" "T20150013886" ...
## $ ABS_CODE : chr "ABS to receive accident" "ABS to receive accident" "ABS to receive accident" "ABS to receive accident" ...
## $ ACCIDENT_STATUS : chr "Finished" "Finished" "Finished" "Finished" ...
## $ ACCIDENT_DATE : chr "2015/07/01 00:00:00+00" "2015/07/01 00:00:00+00" "2015/07/01 00:00:00+00" "2015/07/01 00:00:00+00" ...
## $ ACCIDENT_TIME : chr "1899/12/30 07:00:00+00" "1899/12/30 09:55:00+00" "1899/12/30 21:50:00+00" "1899/12/30 06:00:00+00" ...
## $ ALCOHOLTIME : chr "No" "No" "Yes" "Yes" ...
## $ ACCIDENT_TYPE : chr "Collision with vehicle" "Collision with vehicle" "Collision with a fixed object" "Collision with a fixed object" ...
## $ DAY_OF_WEEK : chr "Wednesday" "Wednesday" "Wednesday" "Wednesday" ...
## $ DCA_CODE : chr "REAR END(VEHICLES IN SAME LANE)" "CROSS TRAFFIC(INTERSECTIONS ONLY)" "OFF RIGHT BEND INTO OBJECT/PARKED VEHICLE" "LEFT OFF CARRIAGEWAY INTO OBJECT/PARKED VEHICLE" ...
## $ HIT_RUN_FLAG : chr "No" "No" "No" "No" ...
## $ LIGHT_CONDITION : chr "Dusk/Dawn" "Day" "Dark No street lights" "Dark Street lights on" ...
## $ POLICE_ATTEND : chr "Yes" "Yes" "Yes" "Yes" ...
## $ ROAD_GEOMETRY : chr "Not at intersection" "Cross intersection" "Not at intersection" "T intersection" ...
## $ SEVERITY : chr "Serious injury accident" "Serious injury accident" "Serious injury accident" "Serious injury accident" ...
## $ SPEED_ZONE : chr "100 km/hr" "60 km/hr" "60 km/hr" "80 km/hr" ...
## $ RUN_OFFROAD : chr "No" "No" "Yes" "Yes" ...
## $ NODE_ID : int 291566 35737 292361 282422 64208 291136 41584 291879 43662 291285 ...
## $ LONGITUDE : num 145 145 145 145 145 ...
## $ LATITUDE : num -38 -38.1 -37.9 -38 -37.8 ...
## $ NODE_TYPE : chr "Non-Intersection" "Intersection" "Non-Intersection" "Intersection" ...
## $ LGA_NAME : chr "CASEY" "CASEY" "YARRA RANGES" "DANDENONG" ...
## $ REGION_NAME : chr "METROPOLITAN SOUTH EAST REGION" "METROPOLITAN SOUTH EAST REGION" "METROPOLITAN SOUTH EAST REGION" "METROPOLITAN SOUTH EAST REGION" ...
## $ VICGRID_X : num 2520841 2526250 2538721 2516548 2485113 ...
## $ VICGRID_Y : num 2392194 2381737 2400595 2389749 2413935 ...
## $ TOTAL_PERSONS : int 4 2 1 1 2 3 4 2 2 2 ...
## $ INJ_OR_FATAL : int 1 1 1 1 1 3 2 1 1 1 ...
## $ FATALITY : int 0 0 0 0 0 1 1 0 0 0 ...
## $ SERIOUSINJURY : int 1 1 1 1 0 2 1 0 1 0 ...
## $ OTHERINJURY : int 0 0 0 0 1 0 0 1 0 1 ...
## $ NONINJURED : int 3 1 0 0 1 0 2 1 1 1 ...
## $ MALES : int 2 2 1 1 1 3 2 2 2 1 ...
## $ FEMALES : int 2 0 0 0 1 0 0 0 0 1 ...
## $ BICYCLIST : int 0 0 0 0 0 0 0 1 0 0 ...
## $ PASSENGER : int 0 0 0 0 0 1 1 0 0 0 ...
## $ DRIVER : int 4 2 1 0 2 2 2 1 1 2 ...
## $ PEDESTRIAN : int 0 0 0 0 0 0 0 0 0 0 ...
## $ PILLION : int 0 0 0 0 0 0 0 0 0 0 ...
## $ MOTORIST : int 0 0 0 1 0 0 1 0 1 0 ...
## $ UNKNOWN : int 0 0 0 0 0 0 0 0 0 0 ...
## $ PED_CYCLIST_5_12 : int 0 0 0 0 0 0 0 0 0 0 ...
## $ PED_CYCLIST_13_18: int 0 0 0 0 0 0 0 0 0 0 ...
## $ OLD_PEDESTRIAN : int 0 0 0 0 0 0 0 0 0 0 ...
## $ OLD_DRIVER : int 0 1 0 0 0 0 0 0 0 0 ...
## $ YOUNG_DRIVER : int 2 0 0 0 0 1 0 0 0 0 ...
## $ ALCOHOL_RELATED : chr "No" "No" "No" "No" ...
## $ UNLICENCSED : int 0 0 0 0 0 0 0 0 0 0 ...
## $ NO_OF_VEHICLES : int 4 2 1 1 2 2 3 2 2 2 ...
## $ HEAVYVEHICLE : int 0 0 0 0 0 0 0 0 0 0 ...
## $ PASSENGERVEHICLE : int 4 1 1 0 2 2 2 1 1 2 ...
## $ MOTORCYCLE : int 0 0 0 1 0 0 1 0 1 0 ...
## $ PUBLICVEHICLE : int 0 1 0 0 0 0 0 0 0 0 ...
## $ DEG_URBAN_NAME : chr "MELB_URBAN" "MELB_URBAN" "MELB_URBAN" "MELB_URBAN" ...
## $ DEG_URBAN_ALL : chr "MELB_URBAN" "MELB_URBAN" "MELB_URBAN" "MELB_URBAN" ...
## $ LGA_NAME_ALL : chr "CASEY" "CASEY" "YARRA RANGES" "DANDENONG" ...
## $ REGION_NAME_ALL : chr "METROPOLITAN SOUTH EAST REGION" "METROPOLITAN SOUTH EAST REGION" "METROPOLITAN SOUTH EAST REGION" "METROPOLITAN SOUTH EAST REGION" ...
## $ SRNS : chr "M" "C" "C" "" ...
## $ SRNS_ALL : chr "M" "C" "C" "" ...
## $ RMA : chr "Freeway" "Arterial Other" "Arterial Other" "Arterial Highway" ...
## $ RMA_ALL : chr "Freeway" "Arterial Other,Local Road" "Arterial Other" "Local Road,Arterial Highway" ...
## $ DIVIDED : chr "Divided" "Divided" "Undivided" "Divided" ...
## $ DIVIDED_ALL : chr "Divided" "Divided,Undivided" "Undivided" "Undivided,Divided" ...
## $ STAT_DIV_NAME : chr "Metro" "Metro" "Metro" "Metro" ...
Data Cont.
As we have explained from the previous slide, we will then subset and
filter the data into a new data called ‘crashes2’. From here, we can see
that the column names have to be changed to make it more easier to read,
and the date has to be changed to only show the year and month. Besides
that, we will have to make the severity into a factor.
crashes2<-crashes[ ,c('ACCIDENT_DATE','SEVERITY','FATALITY','ALCOHOL_RELATED')]
str(crashes2)
## 'data.frame': 75320 obs. of 4 variables:
## $ ACCIDENT_DATE : chr "2015/07/01 00:00:00+00" "2015/07/01 00:00:00+00" "2015/07/01 00:00:00+00" "2015/07/01 00:00:00+00" ...
## $ SEVERITY : chr "Serious injury accident" "Serious injury accident" "Serious injury accident" "Serious injury accident" ...
## $ FATALITY : int 0 0 0 0 0 1 1 0 0 0 ...
## $ ALCOHOL_RELATED: chr "No" "No" "No" "No" ...
colnames(crashes2)<-c("Year","Severity","Fatality","Alcohol-Related")
crashes2$Year<-as.numeric(format(as.Date(crashes2$Year, format="%Y/%m/%d"),"%Y"))
crashes2$Severity<-factor(crashes2$Severity,levels = c("Serious injury accident","Other injury accident","Fatal accident","Non injury accident"),labels=c("Serious","Other","Fatal","No"))
Data Cont.
After formatting the correct data type for each of the variables, we
now have to filter out the data and make it so that it only shows
alcohol related accidents. Here, we have also sorted the year column
into an ascending order using the arrange() function.
alcohol.crashes<-crashes2 %>% filter(`Alcohol-Related`=="Yes")
Data Cont.
Besides that. we have also grouped the accidents by year into a new
data called ‘alcohol.crashes2’, and used the summary() function to
provide the data summary for the accidents.
alcohol.crashes2<-alcohol.crashes %>% mutate(Accident=case_when(!is.na(alcohol.crashes$Year)~"1",is.na(alcohol.crashes$Year)~"0"))
alcohol.crashes2 %>% group_by(alcohol.crashes2$Year) %>% summarise_at(vars(-Year),funs(mean(alcohol.crashes2$Accident,na.rm=TRUE)))
crashes2015<-filter(alcohol.crashes2,Year=="2015") %>% group_by(Year) %>% summarise("No of Accidents" = n())
alcohol.crashes2$Accident<-as.numeric(alcohol.crashes2$Accident)
alcohol.crashes3<-alcohol.crashes %>% group_by(Year) %>% summarise("No of Accidents" = n())
summary(alcohol.crashes2)
## Year Severity Fatality Alcohol-Related Accident
## Min. :2015 Serious:1317 Min. :0.00000 Length:2256 Min. :1
## 1st Qu.:2016 Other : 889 1st Qu.:0.00000 Class :character 1st Qu.:1
## Median :2017 Fatal : 50 Median :0.00000 Mode :character Median :1
## Mean :2017 No : 0 Mean :0.02482 Mean :1
## 3rd Qu.:2019 3rd Qu.:0.00000 3rd Qu.:1
## Max. :2020 Max. :2.00000 Max. :1
Descriptive Statistics and Visualisation
Using the ggplot() function, we have made a bar chart showing the
alcohol-related accidents from the years 2015 to 2020. Surprisingly, the
graph has shown a downward trend in accidents with the year 2017 having
the most alcohol related accidents.
However, there was a sharp drop in alcohol-related accidents in 2020,
presumably that factors such as Covid-19 lock downs were implemented at
that time.
alcohol.crashes4<-data.frame(alcohol.crashes3)
ggplot(alcohol.crashes4,aes(x=Year,y=No.of.Accidents))+geom_bar(width=0.3,stat="identity",color="white") + ggtitle("Alcohol-Related Accidents (2015-2020)")
Decsriptive Statistics Cont.
A table containing the summary of the data is as shown below. Here we
can see that the population mean for the number of accidents throughout
the years are 376, with the highest accident count of 500.
alcohol.crashes3 %>% group_by(Year) %>%summarise(Min = min(`No of Accidents`,na.rm = TRUE),
Q1 = quantile(`No of Accidents`,probs = .25,na.rm = TRUE),
Median = median(`No of Accidents`, na.rm = TRUE),
Q3 = quantile(`No of Accidents`,probs = .75,na.rm = TRUE),
Max = max(`No of Accidents`,na.rm = TRUE),
Mean = mean(`No of Accidents`, na.rm = TRUE),
SD = sd(`No of Accidents`, na.rm = TRUE),
n = n(),
Missing = sum(is.na(`No of Accidents`))) -> table1
knitr::kable(table1)
| 2015 |
247 |
247 |
247 |
247 |
247 |
247 |
NA |
1 |
0 |
| 2016 |
481 |
481 |
481 |
481 |
481 |
481 |
NA |
1 |
0 |
| 2017 |
500 |
500 |
500 |
500 |
500 |
500 |
NA |
1 |
0 |
| 2018 |
457 |
457 |
457 |
457 |
457 |
457 |
NA |
1 |
0 |
| 2019 |
430 |
430 |
430 |
430 |
430 |
430 |
NA |
1 |
0 |
| 2020 |
141 |
141 |
141 |
141 |
141 |
141 |
NA |
1 |
0 |
Decsriptive Statistics Cont.
print(summary(alcohol.crashes3))
## Year No of Accidents
## Min. :2015 Min. :141.0
## 1st Qu.:2016 1st Qu.:292.8
## Median :2018 Median :443.5
## Mean :2018 Mean :376.0
## 3rd Qu.:2019 3rd Qu.:475.0
## Max. :2020 Max. :500.0
Hypothesis Testing
In this presentation, we will be using the two tailed, one-sample
t-test as to test on whether or not there is a difference from the
population mean. As we have seen in the descriptive statistics, we will
confirm on whether or not there is a decrease in alcohol-related
accidents. We would assume that the result will show that there is an
decrease in alcohol-related accidents throughout the years. Since the
data is <30, we will have to adjust and use a t-value test.
The null hypothesis is stated as follows:
\[H_0: \mu_1 = 376 \]The alternate
hypothesis is stated as follows:
\[H_A: \mu_1 \lt 376\]
Hypthesis Testing Cont.
In this step, we have to test our hypothesis. We will use the two
tailed, one-sample t-test to look for the mean difference where mu =
376.
t.test(alcohol.crashes4$No.of.Accidents,mu=376,conf.level=.95)
##
## One Sample t-test
##
## data: alcohol.crashes4$No.of.Accidents
## t = 0, df = 5, p-value = 1
## alternative hypothesis: true mean is not equal to 376
## 95 percent confidence interval:
## 221.948 530.052
## sample estimates:
## mean of x
## 376
Hypthesis Testing Cont.
Here we will check whether or not the data follows a normal
distribution. From here we can see that since there is a small sample
size, we cannot determine whether if it follows a normal
distribution.
alcohol.crashes4$No.of.Accidents %>% qqPlot(dist="norm")
Hypthesis Testing Cont.
A two-tailed, one sample t-test was used to determine if the mean
alcohol-related accidents throughout the years were significantly
different from the population mean of 376 crashes. The significance
level of 0.05 was used. The sample mean of alcohol-related accidents was
376 accidents. Since p,1 >0.01, we can reject the null
hypothesis.
Discussion
- The test result showed that there were a difference in
alcohol-related accidents between 2015-2020.
- We can safely say that there was an decrease in alcohol-related
accidents.
- However, there is a chance that the factor when the pandemic hit
where lock downs were imposed, fewer people got into accidents as there
were less vehicles on the road.
- The limitation of this study was that due to the small sample size
when showed in years, there were difficulties in the findings.
- For the next study, I would propose that there would be more sample
sizes based on the year for a more accurate result.