Avenue de lArme, B.

P : 3900 Kigali, Rwanda Tel: +250 252 574696/252 574698, Fax: +250 252 571925/252 571924

Website: www.ur.ac.rw

SCHOOL OF PURE AND APPLIED SCIENCES

DEPARTMENT OF APPLIED MATHEMATICS
STATISTICS OPTION
Final year Student project Report

TOPIC: APPLICATION OF TIME SERIES

ANALYSIS ON THE OCCURRENCE OF
CAR TRAFFIC ACCDENTS IN RWANDA
Names: NIZEYIMANA Charles GS20120364
Supervisor: Mr. Girum Taye

Project report submitted to College of Sciences and Technology

University of Rwanda in partial fulfilment of the requirements for the First degree in applied
mathematics.
ACADEMIC YEAR 2014-2015

Date:15/06/2015

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA

DECLARATION
I, NIZEYIMANA Charles, hereby declare that, except where otherwise indicated, this
document is entirely my own work and has not been submitted in whole or in part to any
other university.

Name of the student: NIZEYIMANA Charles

Signature of the student........................................

Date.........../............../2015

Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA

DEDICATION
With love and gratitude,
I dedicate this project report to:
My Almighty God,
My whole family,
My friends and
My classmates

Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA

ACKNOWLEDGEMENT
This project report has been made possible by the assistance and encouragement of many
people. Although its not possible to mention all of them here by their names, I would like to
particularly thank the following.
It would not have been possible to complete this project report without the guidance of my
Supervisor Mr Girum Taye, his professional advice and the supervision of this work, his
detailed comments and edits were always to the point and influential, I am much thankful to
him.
I would like to express my appreciation to the government of Rwanda (GOR), more specially
the University of Rwanda (UR), particularly the College of Science and Technology (CST),
the department of Applied Mathematics, School of Applied Sciences for having shaped me to
produce this work.
I also owe special debt gratitude to my family and friends.
Special thanks to Rwanda National Police (RNP) especially the department of Road traffic
accidents for their support and considerable collaboration during the project.
Above all the almighty GOD for the grace and mercy surpasses my understanding.

Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA
ABSTRACT
Road traffic accidents are increasing at a high rate and this has raised major concerns
worldwide including Rwanda. Different reports on traffic accident indicate how people
around the world still losing their lives and other get seriously injured due to the traffic
accidents. Based on the outcomes of accident prediction models, different countermeasures
will be implemented to reduce the frequency of car traffic accidents. Accident-forecasting
models were used to monitor the effectiveness of various road safety policies that have been
introduced to minimise accident occurrences.
The general objective of this paper is to identify and apply the appropriate Box Jenkins time
series model to study the trends and patterns of car traffic accidents occurring in Rwanda as
well as makes a five- year forecast of car traffic accidents that will be expected to occur in
Rwanda. A monthly accident data from 2010 to 2014 was obtained from the traffic accident
department of Rwanda National Police. Those data were deeply analysed using statistical
software SPSS and interpreted to see the characteristic behaviours of car traffic accidents
occurrence in Rwanda. Models were subsequently developed and ARIMA (0, 1,1) was
identified as approximately appropriate model. A five year forecast was made using the
appropriate model and results showed that car traffic accidents would continue to increase
especially in December 2019 to 827 in Rwanda.
Moreover this paper contained the comparison of the true mean among monthly car traffic
accidents to determine the most significant year and month in which most car traffic
accidents occurred by using one way ANOVA and the results showed that there is difference
among the true mean on yearly car traffic accidents where the population mean

2014

has

significantly the highest yearly car traffic accidents but no difference among the true mean of
car traffic accident by month. The findings of this study draw attention to the importance of
implementing key road safety measures in order to change the increasing patterns of car
traffic accidents in Rwanda. Therefore the improved and better policies of National road
safety commission should be introduced with much emphasis on publication and education to
ensure maximum reduction in Car traffic accidents. As conclusion Box Jenkins time series
model was the appropriate time series model on car traffic accidents occurring in Rwanda.
Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA

TABLE OF CONTENTS
Contents

DECLARATION......................................................................................................... i
DEDICATION........................................................................................................... ii
ACKNOWLEDGEMENT............................................................................................ iii
ABSTRACT............................................................................................................. iv
TABLE OF CONTENTS............................................................................................. v
DEFINITION OF TERMS.......................................................................................... vi
LIST OF ACRONYMS.............................................................................................. vii
LIST OF FIGURES................................................................................................. viii
LIST OF TABLES..................................................................................................... ix
CHAP1. INTRODUCTION ON ROAD TRAFFIC ACCIDENTS......................................10
1.1. BACKGROUND........................................................................................... 10
1.2. THEORITICAL INTRODUCTION OF TIME SERIES ANALYSIS............................4
1.3.General Objective......................................................................................... 4
1.4. Specific Objectives...................................................................................... 4
1.5. Research Hypothesis................................................................................... 5
1.6. Research Questions..................................................................................... 5
1.7. Problem Statement...................................................................................... 5
CHAPTER 2: LITERATURE REVIEW..........................................................................6
CHAPTER 3: METHODOLOGY..................................................................................8
3. 1. STUDY SITE AND DATA SOURCE..................................................................8
3.1.1. DATA...................................................................................................... 9
3.1.2. TYPE OF DATA TO BE USED....................................................................9
3.1.3. METHOD OF DATA COLLECTION.............................................................9

Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA
3.2. STATISTICAL OR MATHEMATICAL METHOD USED TO FULFILL THE
OBJECTIVES........................................................................................................ 9
3.2.1. Fitting an appropriate time series model...............................................9
3.2.2. Forecasting future value of monthly average traffic accident
occurrence in Rwanda................................................................................... 11
3.2.4. Compare mean of monthly average traffic accidents occurrence in
Rwanda......................................................................................................... 11
CHAPTER 4: RESULT AND ANALYSIS.....................................................................13
4.1. CAR TRAFFIC ACCIDENTS...........................................................................13
4.1.1. FITTING THE APPROPRIATE BOX JENKINS TIME SERIES MODEL OF CAR
TRAFFIC ACCIDENTS OCCURRENCE IN RWANDA............................................20
4.2. FORECASTING FUTURE VALUES OF MONTHLY CAR TRAFFIC ACCIDENTS 27
4.3. COMPARE MEAN OF MONTHLY CAR TRAFFIC ACCIDENTS OCURRENCE IN
RWANDA........................................................................................................ 33
4.3.1. COMPARE MEAN OF YEARLY CAR TRAFFIC ACCIDENTS OCURRENCE IN
RWANDA........................................................................................................ 34
4. 3.2. COMPARE MEAN OF MONTLY CAR TRAFFIC ACCIDENTS OCURRENCE IN
RWANDA........................................................................................................ 37
CHAPTER 5: DISCUSSION, CONCLUSION AND RECOMMANDATION.......................38
5.1. DISCUSSION............................................................................................... 38
5.2. CONCLUSION............................................................................................. 39
5.3. RECOMMANDATION...................................................................................40

DEFINITION OF TERMS
A time series is a sequence of numerical data that usually occurs in uniform intervals over a
period of time such as hourly, daily, weekly, monthly, yearly etc
Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA
Analysis of Variance (ANOVA) is a statistical method used to compare a response variable
to a number of different levels of factors
Autocorrelation: is the correlation between the target variable and lag values for the same
variables. Correlation (often measured as Correlation coefficient) indicates the strength and
direction of a linear relationship between two random variables.
Correlogram: is a graph in which the sample autocorrelation coefficients are plotted against
time
Forecasting the time series: mean extending the historical values into the future where the
measurements are not available yet.
Multiple Comparisons: is overall test tells us whether there is an existence of difference
between the means for the different groups or not but fail to show where the difference lies.
Prediction Interval: is an estimate of an interval in which future observations will fall, with
certain probability, given what has already observed.
Trend: is the long term upward and downward movement of the series due to factors
influencing the mean of the series.
Stationary time series: It is time series where there is no systematic change in mean (no
trend), no systematic change in variance (no trend) and in which strictly periodic variations
have been removed.

LIST OF ACRONYMS
ACF: Autocorrelation Function
ANOVA: Analysis of Variance
AR: Autoregressive
Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA
ARIMA: Autoregressive Integrated Moving Average
ARMA: Autoregressive Moving Average
PACF: Partial Autocorrelation Function
RNP: Rwanda National Police
SE: Standard Error
SPSS: Statistical Package of Social Science

LIST OF FIGURES
Figure 1: Time plot of car traffic accidents from 2010............................................................26
Figure 2: Time plot of car traffic accidents from 2011............................................................27
Figure 3: Time plot of car traffic accidents from 2012............................................................28
Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA
Figure 4: Time plot of car traffic accidents from 2013............................................................29
Figure 5: Time plot of car traffic accidents from 2014............................................................30
Figure 6: Time plot of monthly car traffic accidents occurrence in Rwanda...........................32
Figure 7 : Original data ACF Correlogram for Car traffic accidents......................................33
Figure 8:Original data PACF Correlogram for Car traffic accidents.......................................34
Figure 9: Time plot of the transformed data of car traffic accidents occurrence in Rwanda...35
Figure 10 :Transformed data ACF correlogram for car traffic accidents.................................36
Figure 11: Transformed data PACF correlogram for car traffic accidents...............................37
Figure 12: Original and forecasted for monthly car traffic accidents......................................39
Figure 13: Histogram of forecast error for monthly car traffic accidents................................40
Figure 14: Histogram of forecast error for monthly car traffic accidents................................40
Figure 15: ACF correlogram of forecast error for monthly car traffic accidents.....................41
Figure 16: Time plot of original, Forecast, Upper and Lower limit bounds for monthly car
traffic accidents........................................................................................................................42

LIST OF TABLES
Table 1: The general one way ANOVA table.....................................................................23
Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA
Table 2: Number of car traffic accidents by year and Month..................................................24
Table 5: Coefficient of the model and their corresponding probability.....................................36
Table 6: Table of forecasted Car traffic accidents by Year and Month......................................41
Table 7: One way ANOVA table of yearly car traffic accidents..............................................43
Table 9: Yearly comparison of car traffic accidents occurrence in Rwanda................................45
Table 8: One way ANOVA table of monthly car traffic accidents............................................46

Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA
CHAP1. INTRODUCTION ON ROAD TRAFFIC ACCIDENTS
1.1. BACKGROUND

Accident is an event, occurring suddenly, unexpectedly and inadvertently under unforeseen

circumstances. Road traffic accidents can be defined as an accident that occurred on a way or
street open to public traffic; resulted in one or more persons being killed or injured, and at
least one moving vehicle was involved. Thus, RTA is collisions between vehicles; between
vehicles and pedestrians; between vehicles and animals; or between vehicles and
geographical or architectural obstacles [1].
Road traffic accidents (RTAs) have appeared to be a huge global public health and
development problem killing almost 1.2 million people a year and injuring or disabling
between 20-50 million people more. The statistical profile shows that in 2002, an estimated
1.2 million people were killed and 50 million injured in road-traffic crashes worldwide,
costing the global community about US $ 518 billion and are third leading causes of death
worldwide, behind malaria and HIV/AIDS [2]. The International Federation of Red Cross and
Red Crescent Societies have described the situation as a worsening global disaster destroying
lives and livelihoods, hampering development and leaving millions in greater vulnerability
[3]. If no means are taken, RTAs and its resultant injuries are predicted to become more
serious as a global burden of disease by 2020, particularly in rapidly motorizing countries [4].
In the developing world, current trends in population growth, industrialization and
urbanization are putting heavy pressure on the transport network in general and on road
system in particular. Compared to the cause of death more commonly associated with the
developing countries, deaths from road accidents are by no means insignificant. This reveals
a serious and growing problem, with absolute fatality and casualty figures rising rapidly in
the majority of developing countries and with death rates considerably higher than in the
developed world [5]. Each day, more than 3000 people or around 85% of all road trafficrelated deaths in the world occur in developing countries ( low and middle income countries)
and for 90% of the annual disability adjusted life years (DALYs) lost because of road traffic
injury. Among the developed countries like the United States of America, Japan and the
United Kingdom traffic related death rate (per 100,000 inhabitants per year) stood at 12.3,
3.85 and 3.59 respectively in the year 2009 [6,7,8,9].
Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA
In addition, Road Traffic Injuries (RTIs) accounts for 30 to 86% of the trauma admissions to
hospitals in low income and middle income countries. Reasons for high burden in road
traffic-related deaths and injuries in developing countries are primarily due to growth in
motor vehicle numbers, Poor enforcement of traffic safety regulations, Inadequacy of public
health infrastructure, Poor access to health services, etc in comparison to the developed
nations [10]. Apart from the humanitarian aspect of the problem, traffic accidents and injuries
in these countries become responsible for an annual loss of $ 65 billion to $100 billion
annually and contribute to poverty by causing deaths, injuries, disabilities, grief, lost of
productivity and material damages. These costs include both loss of income and the burden
placed on families to care for their injured relatives.
Rwanda has a high incidence of serious road accidents for example statistics in the month of
March, 2013, indicate that 477 road accidents occurred in the country which left 30 people
dead and about 204 survived with injuries which also lead to the death of at least a person
every day from road accident .The statistics continue showing that about 2118 people die per
year and 25% of overall deaths and injuries is from people aged between 10 and 40 years old.
[11,12, 13] .
According to Traffic Police Unit statistics, about 1135 people have lost their lives from road
accidents since 2009 and 6943 others get injured in the same period. Between January and
August 2014, 1246 traffic accidents were registered in which 257 people died. Last year, road
accidents reduced by about 6.5 percent to 1882 from 2012 recorded in 2010.[14]
A large percentage of injuries go unreported due to lack of a systematic injury information
system. Moreover, the situation in Rwanda is worsening as RTAs and RTIs have been
increasing over the past twenty years. This is partly due to the increase in the number of
vehicles on the road and partly due to the absence of a coordinated official policy to control
the problem. This ever expanding epidemic is likely to take a heavy burden on the quality of
life and socio-economic growth of Rwanda if no means are taken.
Many research studies incorporating different aspect of RTAs have been carried out by
different workers at different times across the globe. However, such epidemiological studies
Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA
from Rwanda are very much scanty. Moreover, to our knowledge till date there is no such
study ever conducted in Rwanda. Therefore, realizing the need to establish baseline
information on RTAs, the recent study was conducted in Kigali city under the jurisdiction of
Rwanda National Police headquarter of traffic accidents department in Nyarugenge district.
This study is a part of a long term project that investigates the factors leading to the road
traffic accidents and how to prevent them.
Despite being a big problem, the road traffic accidents are preventable. Different countries
around the world including Rwanda have established the road safety strategy aiming to
strengthen the road security and fighting against the traffic accidents.
The overall objective of this research is to find and fit the appropriate model to the traffic
accident and to see the most significant month and year to which the most car traffic
accidents occurred in Rwanda. The aim of this study is to determine and forecast after six
years the pattern of car traffic accidents occurrence in Rwanda.
1.2. THEORITICAL INTRODUCTION OF TIME SERIES ANALYSIS
A time-series is a sequence of numerical data that usually occurs in uniform intervals over
a period of time such as hourly, daily, weekly, monthly, yearly etc. Generally, it is a
sequence of observations ordered by a time parameter. To apply time series analysis, each
time must be associated to a value which is the case in this research because this study
will be based on analysis of monthly average of traffic accidents occurrence in Rwanda
through time. The special characteristics of time series analysis are the fact that
successive observations are usually dependent and that the analysis must take into
account the time order of observations. There are two steps of time series analysis:

Building a model that represent a time series and

Using the model to predict the future value.

When the successive observations are dependent, future values should be predicted from past
observations.
1.3.General Objective
Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA
The general objective of this study is to identify the appropriate Box Jenkins time series
model on the occurrence of car traffic accidents in Rwanda
1.4. Specific Objectives

To fit the appropriate Box Jenkins time series model on the occurrence of car
traffic accidents in Rwanda.
To forecast the future value of the car traffic accidents occurrence in Rwanda.
To compare the means of monthly car traffic accidents occurring in Rwanda by
year and month.
1.5. Research Hypothesis

H 0 : There is no significance difference among the means of monthly car traffic accidents
in Rwanda by year and month
H 1 : There is a significance difference among the means of monthly car traffic accidents in
Rwanda by year and month
1.6. Research Questions

What is the appropriate Box- Jenkins time series Model for the monthly car traffic
accidents occurrence in Rwanda?
Is there a statistical significant difference among the true means of monthly car traffic
accidents occurrence in Rwanda?

Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA

1.7. Problem Statement

Many years ago, many Rwandan populations lost their lives and others get seriously injured
through the traffic accidents and the problem still persists today in Rwanda. Those traffic
accidents are mostly influenced by three different group of factors some related to
environment ( Ex: Adverse of road conditions, slippery roads, bad visibility,...), to the vehicle
(For examples: Bad manoeuvres, Vehicle defects,...) and other related to the drivers (For
examples: Over speeds, Age, Over drinking, Negligence,...).[15]
Many researchers have shown the importance of road user factor which contribute to 95% of
accidents .Poor road user behaviour exhibited by the drivers in some developed countries
may be due to the lack of knowledge regarding the road safety role and regulations, the
general attitude towards the roads safety matter and the problem of alcohol in other countries.
In this project, the plan will be to identify and fit the appropriate Box Jenkins time series on
the car traffic accidents occurring in Rwanda.
The goal of this study is to create a model that will minimise the error between predicted
value of the target variable and the actual value of the traffic accidents.

Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA

CHAPTER 2: LITERATURE REVIEW

This is the review of past related literature available in the area of this study, which shows in
number of studies carried out by several researchers and different authors. As a result it
reviews the relevant literature resulting from the work of interested researchers

and put

forward in the books, report, and any other published documents that will be available to the
researcher. One of the best ways to understand the causes of road traffic accidents is to
develop various accident prediction models which are capable of identifying significant
factors related to human, vehicle, road infrastructure, land-use, and the environment. For
instance,
In Rwanda between 2002 and 2005, The productive youth aged between 16 and 30 years old
counted 50.7% were mostly likely involved in the traffic accidents and 78.7% of the males
were heavily affected and in addition over 90.2% of the traffic accidents took place in Kigali
city.[16]
In 2011, over 4,293 accidents were registered with 3,446 of them minor and 847 serious.
A total of 277 traffic accidents were reported across the country during the first three months
of the year with 72 percent of them in Kigali, according to statistics in the report.
The traffic police report further cites the main causes of road accidents in the country as over
speeding, reckless, driving and, drunk driving.
In Kenya , between 1971 and 1990 about 80% of the deaths involved the pedestrians with
42% and the passengers with 38% and over three thousand

people are killed in road

accidents per year and the 75% of the productive youth adult involved in traffic accidents.
The statistics also shows that human factors, other factors, Vehicle defects and road
Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA
environment played a major role in causing the accidents with 85.5%, 6.4, 5.1% and 2.9%
respectively. The traffic police recorded those road traffic accidents in Kenya rose from 3562
in 1965 to 14342 in 1998 which led to the increasing of the deaths and injuries from 552 to
2972 corresponding to 300% to 430% respectively [17].
Road traffic accident research has extensively been considered within the framework of time
series analysis
In Ghana, models were subsequently developed for accident cases and ARIMA (0,2,1) was
identified with evident as the best model. A Five- year forecast was made using the ARIMA
(0,2,1) model and it showed that road traffic accident cases would continue to increase.
[18]
In China, ARIMA time series model has been found to be appropriate time series model on
the road traffic accidents and the predicted results from ARIMA model revealed that road
traffic accident cases were going to increase from 107,579 to 401,536 over the next one year
in China [19]
In Malaysia, ARIMA model was used to predict number of road fatalities for the year 2015
and 2020 and the results showed that the predicted fatalities for the year 2015 will be 8,760
and 10,716 in 2020.[20]
In Iran, the Box-Jenkins time series model was an acceptable method for prediction of road
traffic accidents and result shown that17643 deaths were recorded by the end of January
where the prediction death was found to be 18,964 by the end of January in 2011. [21]

Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA

CHAPTER 3: METHODOLOGY
The study has been conducted in a systematic procedure starting from selection of the topic to
final report preparation. The integral part was to identify and collect data to be classified,
analyzed, interpreted and presented in a systematic manner to find the vital points. The
overall process of methodology followed in the study is explained further:
3. 1. STUDY SITE AND DATA SOURCE

The study has been conducted on the car traffic accidents occurrence in Rwanda. The data
expected to be used has been obtained from Rwanda National Police especially in the
department of the traffic accident located at MUHIMA in NYARUGENGE District. Different
documents provided by concerned officers and different circulars, manuals and files of the
organization and also different websites related to the study have been used.
3.1.1. DATA

In order to fulfil the objective of the study a series of the data has been used. Those Data are
monthly number of car traffic accidents occurring in Rwanda with observation of 5 years
(From January 2010 to December 2014).The data of 5 years is small but it helped to fulfil the
goal of study
3.1.2. TYPE OF DATA TO BE USED

The present study has been based on secondary source of data from Police records of the
accident cases in Rwanda. The official records have been available from 2010 to 2014.
3.1.3. METHOD OF DATA COLLECTION

The used data has been obtained from Rwanda National Police in the department of traffic
accident by presenting UR-CST recommendation letter. Since the office has organise the data
Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA
which is compiled in computer. In this research, we extracted and used all information which
was useful to fulfil all listed specific objectives. To reach the study objectives, the department
of statistical bulletin in RNP themselves was used as source of information. The reason of
choosing these departments is the role they play in traffic accidents occurrence in Rwanda
and have the data related to the study. This type of method of data collection is called
secondary data source.
3.2. STATISTICAL OR MATHEMATICAL METHOD USED TO FULFILL THE OBJECTIVES
3.2.1. Fitting an appropriate time series model

For the first specific objective which was to fit the appropriate Box Jenkins time series model
for the occurrence of the car traffic accidents in Rwanda, the main method used was Box
Jenkins modelling approach. Firstly, time plot of series was plotted. Because, the data were
neither stationary in mean nor in variance the difference and log transformations have been
applied to the data to make them stationary in mean and in variance respectively.
By taking the reference on the time sequence plot of the study obtained from SPSS, The
choice of the appropriate Box Jenkins time series model was found to be the Autoregressive
Integrated moving average model of order p,d, and q [ARIMA (p,d,q) model]. The
identification of potential model was done based on patterns of the autocorrelation and partial
autocorrelation. To identify the appropriate Box Jenkins model, autocorrelation and partial
autocorrelation coefficients of the stationary data (after data transformation has been made)
was calculated and then depending on the correlograms of the ACF and PACF which
determined the appropriate model.
The autocorrelation at lag k and the partial autocorrelation at lag m were generally given by
the following formulas respectively:
N k

t 1

t 1

rk ( yt yt )( yt k yt ) / ( yt yt ) 2

Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA

Where

yt

: Time series data

y t +k

y t

: mean of time series data

: lag values of time series data at k=1, 2, 3,...

Looking at ACF and PACF if the autocorrelation function cuts off at lag q while partial
autocorrelation function dies away moving average of the order q should be included in the
model whereas when partial autocorrelation function cuts off at lag p while autocorrelation
function dies away autoregressive model of order p should be also included in the model. If
both ACF and PACF die away this suggest combining both (AR and MA) models. We have
also to take into account the order of differencing both seasonal and non seasonal
differencing, for example if the data has been transformed by using non seasonal differencing
and both ACF and PACF die away, ARIMA (p,d,q) model will be fitted where p:order
autoregressive; q: order of moving average and d: order of differencing.
To choose the order of the model, the appropriate method is the principle of parsimony. This
is a rule to seek the simplest model possible. For example if the model to choose is ARMA
(p,q) using this principle, we have to use small p and q as possible. Starting with p=q=1.
In this way, in fitting the appropriate Box Jenkins time series model this principle and
identification of potential model will have been used.
The interpretation based on the output of the time series model chosen will be necessary to
check the adequateness of the model.
Condition of the adequate model
The residuals have no patterns checked by plotting residuals against time

Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA
The residuals are not correlated checked by examining ACF and PACF of the residuals
3.2.2. Forecasting future value of monthly average traffic accident
occurrence in Rwanda

After selecting and fitting the adequate Box Jenkins time series model, the second specific
objective will be fulfilled by forecasting the future values using the available time series data.
This will be done based on finding the model. Since the obtained model is in the function of
lag values to forecast, we will only substitute the corresponding lag values in the model to get
the future values of monthly average accident occurrence in Rwanda.
3.2.4. Compare mean of monthly average traffic accidents occurrence in
Rwanda

To fill the last specific objective, the statistical method called one way analysis of variance
(ANOVA) will be used. The method of one way analysis of variance is used to compare the
response to a number of levels of factor. In this report one way ANOVA is the appropriate
when there is only one factor which is time with different levels which are yearly and
monthly. We will compare the response variable which is monthly car traffic accidents
occurrence in Rwanda to a number of different years and months. The data will be grouped
firstly into different groups with a number of variables for each. To apply this test, the mean
sum of squares for the car traffic accidents occurrence and for the error should be defined.
The mean sum of squares for the car traffic accidents A will be given by:

This is essentially, just the variance of the difference between the mean of each factor level,
and the overall mean. Where we assume that we have k yearly groups with n i cases in the ith
yearly group, and the mean of the i th group is

Xi

while the overall mean is X

and

MSSA: Mean sum of squares.

The mean sum of squares for the error MSSE will be defined as follows:

Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA

Where Si2 is the sample variance for the ith yearly group
Here the Null hypothesis to be tested is that, there is no difference between the yearly groups.
The Alternative hypothesis is that there is a difference. Note that the alternative hypothesis, in
the situation of comparing several levels of car traffic accident occurrence is always a two
sided alternative, as it does not make sense to try to specify in advance where the differences
lie.
The ratio of the above two mean sum of squares has an F distribution (that is MS A/MSE has F
distribution) with the two degrees of freedom associated with MS A which is k-1 and for MSE
which is N-K. Here the Null hypothesis is rejected if the p-value is smaller than the specified
significance level (or if the test statistic F is larger than the critical value) i.e there is
difference among true mean on yearly car traffic accident occurrence in Rwanda.
Table 1: The general one way ANOVA table is as follows:
Source

of Sum

variation

of d.f

Mean Square

F-Ratio

K-1

SSA /d.f

MSA/MSE

SSE /d.f

Sig. level

squares

Between
Years

and SSA

Month
Within Years
and Month

SSE

N-K

Total

TSS

N-1

Table 1: The general one way ANOVA table

Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA

CHAPTER 4: RESULT AND ANALYSIS

In this chapter, we have all related results to the car traffic accidents occurrence in Rwanda in
the form of Tables and Figures which prompt us to know and analyze the appropriate time
series model of car traffic accidents occurrence in Rwanda. Those results also provide how
the model fitted the data which is helpful in forecasting stage and the result of comparing the
true mean among the means of car traffic accidents by year and Month.
4.1. CAR TRAFFIC ACCIDENTS

Here are the tables which represent monthly and yearly car traffic accidents occurrence in
Rwanda from 2010 to 2014.
Table 2: Number of car traffic accidents by year and Month
Month

Number of accident by Years

Total
2010

2011

2012

2013

2014

January

354

410

347

549

476

February

381

389

396

476

488

March

378

411

398

531

490

April

440

372

379

350

473

May

406

451

451

456

487

2136
2130
2208
2014
2251

Mean

427.2
426
441.6
402.8
450.2

Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA
June

373

364

468

517

440

July

394

370

442

516

562

August

413

353

443

480

588

September

438

344

553

471

539

October

392

330

633

485

507

November

463

320

607

499

474

December

452

329

297

474

596

Total in year

4884

4403

5414

5804

6120

Mean

407

367

452

484

510

2162
2284
2277
2345
2347
2363
2148

432.4
456.8
455.4
469
469.4
472.6
429.6

26625

Table 2: Number of car traffic accidents by year and Month

It is observed that car traffic accidents vary with respect to month and Year.

Detecting the seasonality in the data

Seasonality in a time series data can be identified by regularly spaced peaks and troughs
which have a consistent direction and approximately the same magnitude every year, relative
to the trend. The following time plots depict the peaks of each year from 2010 to 2014.

Figure 1: Time plot of car traffic accidents from 2010

Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

44486000
44320800
3602J0A1N2F0E1B2M
N
Jte
0A1R20P1RM
20A1Y2J0U
1D
2a
0U
1L2A0U
1G
2S0E1P2O
0C
1T2N
0O
1V2D
0E1C

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in

Rwanda

M
onthlycartficaidents

Charles NIZEYIMANA

From the graph displayed above, the highest peak was found in November and the lowest
peak was found in month of January.

Figure 2: Time plot of car traffic accidents from 2011

Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

44200
338600
3342002J0A1N2F0E1B2M
N
Jte
0A1R20P1RM
20A1Y2J0U
1D
2a
0U
1L2A0U
1G
2S0E1P2O
0C
1T2N
0O
1V2D
0E1C

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in

Rwanda

M
onthlycartficaidents

Charles NIZEYIMANA

From the graph displayed above, the highest peak was found in May and the lowest peak was
found in month of November.

Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA

6500
4300
N
Jte
2J0A1N2F0E1B2M
0A1R20P1RM
20A1Y2J0U
1D
2a
0U
1L2A0U
1G
2S0E1P2O
0C
1T2N
0O
1V2D
0E1C

M
onthlycartficaidents

Figure 3: Time plot of car traffic accidents from 2012

From the graph displayed above, the highest peak was found in October and the lowest peak
was found in month of December.

Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA

5500
44500
3502J0A1N
N
32F0E1B32M
0A1R
320P1R
3M
20A1Y32J0U
1D
3a
2tJe
0U
1L32A0U
1G
32S0E1P32O
0C
1T32N
0O
1V32D
0E1C
3

M
onthlycartficaidents

Figure 4: Time plot of car traffic accidents from 2013

From the graph displayed above, the highest peak was found in January and the lowest peak
was found in month of April

Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda

6500
54050
402J0A1N
N
42F0E1B42M
0A1R
420P1R
4M
20A1Y42J0U
1D
4a
2tJe
0U
1L42A0U
1G
42S0E1P42O
0C
1T42N
0O
1V42D
0E1C
4

Charles NIZEYIMANA

M
onthlycartficaidents

Figure 5: Time plot of car traffic accidents from 2014

From the graph displayed above, the highest peak was found in December and the lowest
peak was found in month of June.

Since there are no regularly spaced peaks and through which have a consistent direction and
approximately the same magnitude every year, relative to the trend, this implies that there is
no seasonality in the data and the absence of seasonality implies that SARIMA models can
cannot be identified as the appropriate time series model to fit the data .The presence of
trends in the data suggests that simple exponential smoothing models also cannot be
identified as the appropriate time series model to fit the data therefore it remains only to
identify the appropriate time series model among the ARIMA models.
Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA

4.1.1. FITTING THE APPROPRIATE BOX JENKINS TIME SERIES MODEL OF

CAR TRAFFIC ACCIDENTS OCCURRENCE IN RWANDA

To fit the appropriate Box Jenkins time series model and to be able to forecast the future
values for car traffic accidents occurrence in Rwanda the following procedure was used. Time
series data was plotted in the form of time plot before applying any transformation which

6500
4300
Yte
2J0A1NM
20A1Y2S0E1P2J0A1NM
20A1Y2S0E1P2JA01NM
2D
0A1a
2S0E1P2JA01N
3M
20A1Y32S0E1P32J0A1N
4M
20A1Y42S0E1P4

helped to check if the data are stationary in mean and in variance. The correlogram or plots of
autocorrelation function (ACF) and the partial autocorrelation function (PACF) were plotted
below:

M
onthlycartficaidents

Time plot below shows original data of car traffic accidents occurrence in Rwanda

Figure 6: Time plot of monthly car traffic accidents occurrence in Rwanda

Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA
As it is seen on above figure 6, the data is neither stationary in mean nor in variance because
the time plots above shows a systematic change in mean (trend) as well as in variance. Since
Box Jenkins cannot be fitted to the data which are not stationary, the data must be
transformed firstly to make them stationary in mean and in variance by applying log and
difference transformations respectively.
ACF figure below shows original data of monthly car traffic accidents occurrence in Rwanda
Monthly car traffic accidents.

Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA

Coefficient

Confidence Limit

Upper

Confidence Limit

Lower

Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Monthly
carSeries
trafficon
accidents
Application
of Time
Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA
Figure 7 : Original data ACF Correlogram for Car traffic accidents
From the above ACF figure7, it shows pattern, does not decay to zero which indicates that the
original data are neither stationary in mean nor in variance and this conduct to the log and
difference transformations of the data to make both mean and variance constant respectively.

PACF figure below shows original data of monthly car traffic accidents occurrence in
Rwanda

0.5
0.0

1.0
-0.5

-1.0

ACF

Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application
Series
on
1 2 3 of
4 5 Time
6 7 8 9 10
1112131415
16 Car Traffic Accidents in
Rwanda
Lag Number
Charles NIZEYIMANA

Monthly car traffic Accidents

1.0

Coefficient
0.5

Part
ial
ACF

Lower
Limit
Confidence
Upper Confidence
Limit

0.0

-0.5

-1.0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

Lag Number
Figure 8: Original data PACF Correlogram for Car traffic accidents
According to the above PACF figure correlogram ,it shows that ARIMA (Autoregressive
Integrated Moving Average) model can be identified as appropriate model. This model is
classified as an ARIMA (p,d,q) where p,d,q represent a non seasonal order of autoregressive,
differencing and moving average respectively. To fit this model, we are required first to make
data stationary in mean and in variance by appropriate transformations which are log
transformation to make variance stationary and difference transformation to make mean

Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA
stationary. Time plot of the data after transformation and their corresponding ACF and PACF

000...63
--00..36FEBJUNOCTFEBJUNOCTFEBJUNOCTFEBJUNOCTFEBJUNOCT
2012012T
01ra
2n
0s
1fo
2r0m
1:20n
1a
2tu
0r1lD
20o
1a
2g
0te
1,d
2if0r13n
2c
0e
13(1
2)013201420142014

are displayed below:

Time plot below shows stationary of car traffic accidents occurrence in Rwanda after

M
onthlycartficaidents

transformations.

Figure 9: Time plot of the transformed data of car traffic accidents occurrence in
Rwanda
The above plot 9 is the time plot of the transformed data as it can be seen the data become
stationary with respect to mean and variance by taking difference of order 1 and log
transformations.

Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA
ACF figure below shows stationary of car traffic accidents after transformation

Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

M
o
n
t
h
l
y
c
a
r
t
f
i
c
a
i
d
e
n
t
s
oLC
eow
fnfidircncteLim
10..05 C
U
p
t
0.

Application of Time Series on Car Traffic Accidents in

Rwanda
Charles NIZEYIMANA
Monthly car traffic accidents

Figure 10 : Transformed data ACF correlogram for car traffic accidents

Looking at figure 10, it shows that ACF decay to zero which indicates the stationality of the
data in mean and in variance by taking difference of order 1 and log transformation.

PartilA
C
F

PACF figure below shows correlation of car traffic accidents after transformation.

--01..50123456L
7a
8g
9N
1u
0m
1
2
3
1
4
5
1
6
ber

Confidence

Coefficient

0.511: Transformed data PACF correlogram for car traffic accidents

Figure
0.0

According to the figure11displayed above, the correlogram shows a high negative correlation
1.0 the current value and their value at lag 3.
between
-0.5

-1.0

Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
ACF

Charles NIZEYIMANA

4.1.1.1. FITTING THE APPROPRIATE BOX JENKINS TIME SERIES MODEL OF

MONTHLY CAR TRAFFIC ACCIDENTS

From above figures

correlograms,
the sample ACF shows a
1 2 3104and
5 611,7 ACF
8 9 and
10 11PACF
12 13 14
15 16

Lag
significant cut off than PACF
andNumber
both show the same negative value at lag1 and much
smaller values at high lags for ACF. Based on above information and using the principle of
Parsimony which is seeking the simplest model as much as possible; the appropriate model to
be fitted is ARIMA (0,1,1) which is moving average of order 1 and differencing of order 1.
Its general equation is given by the following formula:
y t = + t 1 + t1

Where

: Coeffient of MA (1)

: Error term

: Constant

Therefore, fitting ARIMA model for our data using SPSS by letting p=0, d=1 and q=1 give
us the following output:
Table 5.Coefficient of the model and their corresponding probability
ARIMA Model Parameters
Accidents-Model_1 Accidents Natural Log

Constant
Difference
MA

Lag 1

Estimate
.006
1
.695

SE
.006

t
1.060

Sig.
.294

.098

7.115

.000

Table 3: Coefficient of the model and their corresponding probability

The coefficient of MA (1) is 0. 695 and its corresponding probability is 0.000.The coefficient
is significant at 95% confidence interval because its probability is less than level of
Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA
significant of 5%. Thus the actual fitted model based on the above estimated coefficient is as
follow:
y t =0.0060.695 t1 + t

4.2.

FORECASTING

FUTURE

VALUES

OF

MONTHLY

CAR

TRAFFIC

ACCIDENTS

After obtaining the appropriate Box Jenkins time series model fitting good the data, we
proceed to forecast the future values of car traffic accidents in coming five years. The
adequateness of the model was checked before forecasting since forecasting cant be made
without checking if the model fitted is good or not. Prediction of future values is made to a
model which fist good the data and the goodness of fitted model were showed using ACF and
Histogram super imposed by normal curve of the error terms. To check if the fitted values are
close to the original data, multiple time plots ware plotted below:
Multiple time plots below shows fitted values and monthly car traffic accidents

Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA

600

500

400

300

Monthly car traffic Accidents

Forecasted
JANAPRJULOCTJANAPRJULOCTJANAPRJULOCTJANAPRJULOCTJANAPRJULOCT
2010
2010
2010
2010
2011
2011
2011
2011
2012
2012
2012
2012
2013
2013
2013
2013
2014
2014
2014
2014

Date
Figure 12: Original and forecasted for monthly car traffic accidents
From the above time plot figure12, original data and forecasted values move together which
implies that the fitted values match the observed values closely. This shows that the model
fitted is good.

Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA

Histogram of forecasts error for monthly car traffic accidents occurrence in Rwanda.
Figure 13: Histogram of forecast error for monthly car traffic accidents
20

15

10

Mean =0

Std. Dev. =0. 137

N =59

0
-0.75

-0.5

-0.25

0.25

Error from Monthly car traffic

Accidents-Model_1
Figure 14: Histogram of forecast error for monthly car traffic accidents
The Histogram above super imposed by normal curve on monthly data indicates that the
curve is approximately normally distributed with mean zero and the standard deviation of
0.137.
The ACF of forecast error for monthly car traffic accidents.

Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA

Error from Monthly car traffic accidents

Coefficient

Confidence Limit

Upper

Confidence Limit

Lower
Application of Time series on Car traffic
accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA
Figure 15: ACF correlogram of forecast error for monthly car traffic accidents
From the figure 15, it is observed that the error is approximately uncorrelated because all
coefficients except at lag 3 lie approximately inside the boundaries especially at highest lag.

0.5
0.0

Multiple time plots showing original, Forecast, Upper and lower limit bounds for
1.0

-0.5 car traffic accidents

monthly

ACF

-1.0

Application of Time series on Car traffic accident occurrence in Rwanda

1 2 3 4 5 6 7 8 9 10111213141516

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA
Monthly car
traffic accidents
Forecasted
LCL from
AccidentsModel_1
UCL from
AccidentsModel_1

1,500

1,250

1,000

750

500

250

JAN NOVSEP JUL MAYMARJAN NOVSEP JUL MAYMAR

2010
201020112012
20132014
201520152016
201720182019

Date
Figure 16: Time plot of original, Forecast, Upper and Lower limit bounds for monthly
car traffic accidents
Looking at the figure 16, the original and forecast data move together but with increasing and
slightly decreasing of Upper and Lower limit bounds respectively from 2015 due to the lack
of observations from 2015 up to 2019.In this range, the plot of values are the forecasted
values. The prediction interval for forecasts is increasing (wide) for the specified period in
which the data are available (From January 2010 to December2014) and it decreases for the
specified period in which the data are not available (From January 2015 to December2019).
By looking the output of the ACF of forecast error term, all coefficients of the error term at
different lags lie approximately inside the boundaries except at lag 3, indicates the
uncorrelation between the errors. The error terms mean are close to zero with constant
variance, the super imposed normal curves for error terms are approximately normal.
Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA
Multiple time series plot shows that the fitted values and the original data move and close
together which describe the adequate model. We proceed to forecast in five years the number
of monthly car traffic accidents especially in December since the model are accepted.
Monthly car traffic accidents in next 5years especially in December will be 827 and their
corresponding prediction interval is (383, 1575).
Table 4: Table of forecasted Car traffic accidents by Year and Month
Monthly
Forecaste

Forecasted in years
2015
2016
2017

2018

2019

d
January
February
March
April
May
June
July
August
September

548
552
556
560
564
567
571
575
579

596
600
604
608
613
617
621
626
630

648
652
657
662
666
671
675
680
685

704
709
714
719
724
729
734
740
745

766
771
777
782
787
793
798
804
810

October
November

584
588

634
639

690
695

750
755

815
821

December

592

643

699

760

827

Compared to the five years ago, Car traffic accidents will increase over the next five years.

Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA
4.3. COMPARE MEAN OF MONTHLY CAR TRAFFIC ACCIDENTS OCURRENCE
IN RWANDA.

To fulfil the specific objective which is to compare the true mean of car traffic accidents
occurrence in Rwanda ,the technique of one way analysis of variance (One way ANOVA)
was used through year 2010 to 2014.
Time was identified as the independent variable which is a single factor with five groups
(yearly) and with twelve groups (monthly) and the dependent variable was found to be
monthly car traffic accidents occurrence in Rwanda. One way ANOVA was used since we are
dealing with one factor with different group. The mean sum of squares for factor A is given
by the following formula:

This is the variance of the difference between the mean of each factor level, and the
overall mean. We assume that we have k groups with

ni ( i=1,2, ,5 ) cases in the ith group,

and the mean of the ith group is X i , while the overall mean is X

in comparison of

monthly car traffic accidents occurrence in Rwanda. The mean sum of squares for the error
MSSE is defined as follows:

Where Si2 is the sample variance for the ith yearly groups
The ratio of the above two mean sum of squares has an F distribution (that is MS A/MSE has F
distribution) with the two degrees of freedom associated with MS A which is k-1 and for MSE
which is N-K, where N is the overall observations (grand total). Here the Null hypothesis is
Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA
rejected if the p-value is smaller than the specified significance level (or if the test statistic F
is larger than the critical value) i.e there is difference among true mean on yearly car traffic
accident occurrence in Rwanda.
4.3.1. COMPARE MEAN OF YEARLY CAR TRAFFIC ACCIDENTS OCURRENCE
IN RWANDA.

From table 2(Monthly car traffic accidents) there is hypothesis to be tested. This hypothesis
must be tested using one way ANOVA since there is only one factor with different levels.
Those hypotheses are the following:
H 0 : 2010=2011 = 2012=2013 =2014

(There is no difference among the true mean on yearly

car traffic accidents)

H 0 : 2010 2011 2012 2013 2014

(There is difference among the true mean on yearly car

traffic accidents)

Where

2010

2011

2012

2013

2014

represent the population mean of yearly car

traffic accidents. The result of one way ANOVA analysis helped in making conclusion. Using
SPSS the result obtained are summarized in one way ANOVA table as follow:
Table 7.One way ANOVA table of yearly car traffic accidents
ANOVA
Accidents
Source

of Sum

of

Mean

variation
Squares
d.f
Square
F
Between years 159496.000
4
39874.000
11.265
Within years
194683.250
55
3539.695
Total
354179.250
59
Table 5: One way ANOVA table of yearly car traffic accidents

Sig.
.000

Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA

The result above allows us to test whether there is a significant difference in population
mean of monthly car traffic accidents or not. The p-value (sig.) in the output which is 0.000 is
smaller than level of significance which is 5%. Therefore, this suggests rejecting the null
hypothesis (H0). In rejecting the null hypothesis it means that there is significant difference
in yearly car traffic accidents. By comparing the value of F-calculated which is 11.265 and
the value of F from the table with degree of freedom (k-1, N-K) =(4,55) which is 2.54 we
obtain that F-calculated > F-tabulated, therefore, we reject H0 which is the same conclusion
as that obtained using P-value (that there is significant difference in yearly car traffic
accidents).

4.3.1.1. MULTIPLE COMPARISONS

Since there is difference among the true mean on yearly car traffic accidents occurrence in
Rwanda, generally, we would like to know which means are significantly different. This was
done by comparing all pair wises between pairs of the year. We determine whether or not zero
is included in the confidence intervals. Tukeys method was used for all pair wise
comparisons since it is commonly used procedure which is more powerful than Bonferroni
method. The general form, allowing for different sample sizes for treatment i and j is:

(Yi Yj ) q , k , n k MSE (1/ ni 1/ n j ) / 2

Where

q , k ,nk

is the studentized range from the table

The difference among the means of car traffic accidents is shown in the following:

Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA

Table9.Yearly comparison of car traffic accidents occurrence in Rwanda

YEARLY

DIFFERENC

P-

95%CONFIDENC

CONCLUSIO

COMPARISON

VALUE

E INTERVAL

MEAN

LOWER UPPER
BOUND BOUND

2010-2011

40.083

0.473

-28.42

108.59

2010-2012

-44.167

0.374

-112.67

24.34

2010-2013

-76.667

0.021

-145.17

-8.16

2010-2014

-103.000

0.001

-171.50

-34.50

2011-2012

-84.250

0.009

-152.75

-15.75

2011-2013

-116.750

0.000

-185.25

-48.25

2011-2014

-143.083

0.000

-211.34

-74.75

2012-2013

-32.5

0.669

-101.00

36.00

2013-2014

-26.333

0.814

-94.84

42.17

2010 = 2011
2010 = 2012
2010 < 2013
2010 < 2014
2011 < 2012
2011 < 2013
2011 < 2014
2012= 2013
2013 = 2014

Table 6: Yearly comparison of car traffic accidents occurrence in Rwanda

Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA
According to the above table, at 95% confidence interval
2010 = 2011=2012 < 2013 =2014

2014

; We conclude that the population mean

has

significantly the highest yearly car traffic accidents.

4.

3.2.

COMPARE

MEAN

OF

MONTLY

CAR

TRAFFIC

ACCIDENTS

OCURRENCE IN RWANDA

From table 4 (Monthly car traffic accidents) there is hypothesis to be tested. This hypothesis
must be tested using one way ANOVA since there is only one factor with different levels.
Those hypotheses are the following:
H 0 : Jan= Feb= Mar 2013= Apr May= Jun= July= Aug = Sep= Oct= Nov = Dec

i.e. there

is no difference among the means of monthly car traffic accidents in Rwanda

H 1 : Jan Feb Mar Apr May Jun July Aug Sep Oct Nov Dec

Or Not all

(i=January... December) is zero.

(There is difference among the means of monthly car traffic accidents)
Where

, i= January... December represent the population mean of monthly car traffic

accidents. The result of one way ANOVA analysis helped in making conclusion. Using SPSS
the result obtained are summarized in one way ANOVA table as follow:

Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA
Table 8.One way ANOVA table of monthly car traffic accidents
ANOVA
Source

of

variation
Between

Sum of Squares d.f

Mean Square

25316.133
11
2301.467
.336
Month
Within Month 329157.600
48
6857.450
Total
354473.733
59
Table 7: One way ANOVA table of monthly car traffic accidents

Sig.
.973

The result above, the p-value (sig) is .973 which is greater than level of significance of
5%.Therefore; this means the acceptance of the null hypothesis no significant difference in
population mean of monthly car traffic accidents by month. In accepting the null hypothesis it
means that there is no significance difference in monthly car traffic accidents. By comparing
the value of F-calculated which is 0.335 and the value of F from the table with degree of
freedom (k-1, N-K) =(11,48) which is .... we obtain that F-calculated < F-tabulated, therefore,
we accept H0 which is the same conclusion as that obtained using P-value (that there is no
significant difference in mean of monthly car traffic accidents

CHAPTER

5:

DISCUSSION,

CONCLUSION

AND

RECOMMANDATION
5.1. DISCUSSION

This section discussed the main results obtained in previous chapter and compares those
results with results obtained by others worldwide. The data analyzed are monthly car traffic
accidents occurrence in Rwanda. During the analysis, the appropriate box Jenkins time series
model was found to be ARIMA (0,1,1).Using the ARIMA models, we forecasted the future
value of car traffic accidents of the next 5 years whose results are shown in Figures16.
The prediction results are accepted and 95% of the confidence interval of ARIMA is wider. In
brief it will help both victims and the drivers to understand well the dangers of car traffic
accidents to the life of population and generally to the economic development of Rwanda.
Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA
Monthly Car traffic accidents in the next 5years especially in December 2019 will be 827 and
was found using fitted model. The yearly comparison of car traffic accident occurrence in
Rwanda using one way anova showed a significant difference among the means of yearly car
traffic accidents in Rwanda and using Tukeys method for all pair wise comparisons were
found that the population mean

2014

has significantly the highest yearly car traffic

accidents.
In comparison with other related study, the result showed that generally ARIMA time series
model has been found to be appropriate time series model on the car traffic accidents and the
predicted results from ARIMA model worldwide revealed that car traffic accident cases were
going to increase over the next six year from ARIMA (0,1,1) in Rwanda, 107,579 to 401,536
over the next one year in China[19] and also in Ghana, ARIMA (0,2,1) was identified with
evident as the best model. A six- year forecast was made using the ARIMA (0, 2, 1) model
and it showed that car traffic accident cases would continue to increase. [18]
5.2. CONCLUSION

Time series Analysis of the car traffic accidents data from January 2010 to December 2014
showed that the patterns of car traffic accidents are increasing in Rwanda. ARIMA (0,1,1) is
identified to be approximately the suitable model for car traffic data. Prediction from ARIMA
(0, 1, 1) model revealed that car traffic accidents will continue to increase over the next five
years. The findings of this study draw attention to the importance of implementing key road
safety measures in order to change the increasing patterns of car traffic accidents in Rwanda.
Therefore the improved and better policies of National road safety commission should be
introduced with much emphasis on publication and education to ensure maximum reduction
in Car traffic accidents.

Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA

5.3. RECOMMANDATION.

Depending on result obtained, this study highlights recommendations which include the
advices to Rwanda National Police (RNP).
To minimize the occurrence of the car traffic accidents, the following measures would be
taken as recommendations:
To educate the public road user especially the drivers and pedestrians on how to abide
by the low in order to reduce traffic accidents.
To implement and improve road safety campaign
Sensitize the road users to use correctly the established public parking yard and
increase in number of public parking yard following a complaint by taxi operators
that the yards available are insufficient

Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA
To enforce the rules and lows to ensure security and association of the transporters
should participate to formalize penalties to members who violate traffic rules and
regulation as a way of enhancing road security
Regular inspection of vehicles by MIC (Motor vehicle inspection center) and
technical control certificate issued after the inspection
The marking of dividers and speed breaker must be done. Immediate to remove all the
aged and damaged vehicles off from the road will definitely be a welcoming step.
Government and NGOs should create public awareness to prevent road traffic
accident by organizing seminars, road drama etc

REFERENCES
1. The Oxford Advance Learners Dictionary (1989)
2. Peden M, Scurfield R, Sleet D, Mohan D, Hyder AA, Jarawan E (Eds.) 2004. World
report on road traffic injury prevention. Geneva: World Health Organization.
3. Cater N, Walker P (Eds.) 1998. World Disasters Report. Oxford University Press.
4. Murray CJL, Lopez AD (Eds.) 1996. The global burden of disease: a comprehensive
Assessment of mortality and disability from diseases, injuries, and risk factors in 1990
And projected to 2020. Cambridge, MA: Harvard School of Public Health on behalf of the
World Health Organization and the World Bank.
Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA
5. World Health Organization 1984. Road traffic accidents in developing countries.
Technical Report Series No. 73. Geneva.
6. Sleet DA, Branche CM 2004. Road safety: A new public health priority. Safety Science
Monitor, 1: 1-3.
7. "IRTAD Database, November 2009 - Risk Indicators". OECD International Traffic Safety
Data and Analysis group (IRTAD).
[http://www.internationaltransportforum.org/irtad/pdf/risk.pdf.
8. http://www.japantoday.com/category/national/view/traffic-accident-deaths-hit-57-yearlowbelow-5000-in-2009
9. http://www.dft.gov.uk/pgr/statistics/datatablespublications/accidents/casualtiesmr/
rrcgbmainresults2009
10. Nantulya VM, Reich MR 2002. The neglected epidemic: road traffic injuries in
developing countries. BMJ, 324: 1139-1141.
11.New times, Standard Journal,Vol.1,Number 10
12. World Health Organization global status report on road safety 2013
13. WHO-Rwanda burden of disease 2010 report
14. Police magazine no 005.
15. Volvo Trucks, European Accident Research and Safety Report (2013)

16: East and Central African Journal of Surgery, Vol. 13, No. 1, March-April 2008, pp. 7376 , E. Twagirayezu, R. Teteli, A. Bonane, E. Rugwizangoga, Kigali University Teaching
Hospital, Kigali, Rwanda
Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA
17. Injury Control and Safety Promotion (2003), Vol.10, No.1-2, pp53-61]
18.Avuglah R., Adu-Poku K.A, Harris E (2014). The application of ARIMA models to road
traffic accident case in Ghana].
19. Yuan-Yuan Pack et al., 2013)
Yuan, Y.P., Xu- Jun, Z., Zhi-bin, T., Meng-jing, C., and Yue, G. (2013). Autoregressive
Integrated Moving Average Model in predicting road traffic injury in China. Zhonghua Liu
Xing Bing Xue Za Zhi 2013 Jul; 34(7):736-9.
20. Rohayu Sarani et al., 2012). Rohayu, S., Sharifah Allyana, S.M.R., Jamilah, M.M., and
Wong, S.V. (2012). Predicting Malaysian road fatalities for year 2020. Kuala Lumpur:
Malaysian Institute of Road Safety Research 2020, MRR(6)2012
21.Journal of Asian Scientific Research, 2013, 3(4):425-430

Application of Time series on Car traffic accident occurrence in Rwanda

Page 1

Research Project |University of Rwanda

Application of Time Series on Car Traffic Accidents in
Rwanda
Charles NIZEYIMANA

Application of Time series on Car traffic accident occurrence in Rwanda

Page 1