CHAPTER ONE

INTRODUCTION

1.1 BACKGROND OF THE STUDY

Time series modeling is a dynamic research area which has attracted attentions of researchers.

in literature community over last few decades. The main aim of time series modeling is to

carefully collect and study the past observations of a time series to develop an appropriate model

which describes the inherent structure of the series. This model is then used to generate future

values for the series, i.e. to make forecasts. Time series forecasting thus can be termed as the act

of predicting the future by understanding the past (Raicharoen 2003). Due to the indispensable

importance of time series forecasting in numerous practical fields such as business, economics,

finance, science and engineering, etc. (Zhang et al; 2007), proper care should be taken to fit an

adequate model to the underlying time series. It is obvious that a successful time series

forecasting depends on an appropriate model fitting. A lot of efforts have been done by

researchers over many years for the development of efficient models to improve the forecasting

accuracy. As a result, various important time series forecasting model has been evolved in

literature.

In time series analysis, the moving-average model (MA model), also known as moving-

average process, is a common approach for modeling univariate time series. The moving-average

model specifies that the output variable depends linearly on the current and various past values

of a stochastic (imperfectly predictable) term. Despite, The moving-average model should not be

confused with the moving average, a distinct concept some similarities, Contrary to the AR

model, the finite MA model is always stationary. The notation MA(q) refers to the moving
average model of order q: μ is the mean of the series, the θ1, ..., θq are the parameters of the

model and the εt, εt−1,..., εt−q are white noise error terms. The value of q is called the order of the

MA model. This can be equivalently written in terms of the backshift operator B as, Thus, a

moving-average where model is conceptually a linear regression of the current value of the series

against current and previous (observed) white noise error terms or random shocks. The shocks at

each point are assumed to be mutually random independent and to come from the same

distribution, typically a normal distribution with location at zero and constant scale.

In statistics a moving average (rolling average or running average) is a calculation to

analyze data points by creating a series of averages of different subsets of the full data set. It is

also called a moving mean (MM) or rolling mean and is a type of finite impulse response filter.

Variations include: simple and cumulative or weighted forms (described below). Given a series

of numbers and a fixed subset size, the first element of the moving average is obtained by taking

the average of the initial fixed subset of the number series. Then the subset is modified by

"shifting forward"; that is, excluding the first number of the series and including the next value

in the subset. A moving average is commonly used with time series data to smooth out short-

term fluctuations and highlight longer-term trends or cycles. The threshold between short-term

and long-term depends on the application, and the parameters of the moving average will be set

accordingly. For example, it is often used in technical analysis of financial data, like stock prices

return or trading volumes. It is also used in economics to examine gross domestic product,

employment or other macroeconomic time series. Mathematically, a moving average is a type of

convolution and so it can be viewed as an example of a low-pass filter used in signal processing.

When used with non-time series data, a moving average filters higher frequency components
without any specific connection to time, although typically some kind of ordering is implied.

Viewed simplistically it can be regarded as smoothing the data.

Time series is a sequential set of data points, measured typically over successive times. It

is mathematically defined as a set of vectors x(t),t = 0,1,2,... where t represents the time elapsed

(Cochrane et al 1997). The variable x(t) is treated as a random variable. The measurements taken

during an event in a time series are arranged in a proper chronological order. Time series

containing records of a single variable is termed as univariate. But if records of more than one

variable are considered, it is termed as multivariate. A time series can be continuous or discrete.

In a continuous time series observations are measured at every instance of time, whereas a

discrete time series contains observations measured at discrete points of time. For example

temperature readings, flow of a river, concentration of a chemical process etc. can be recorded as

a continuous time series. On the other hand population of a particular city, production of a

company, exchange rates between two different currencies may represent discrete time series.

Usually in a discrete time series the consecutive observations are recorded at equally spaced time

intervals such as hourly, daily, weekly, monthly or yearly time separations. As mentioned in

(Hipel et al 1994), the variable being observed in a discrete time series is assumed to be

measured as a continuous variable using the real number scale. Furthermore a continuous time

series can be easily transformed to a discrete one by merging data together over a specified time

interval.

Time series in general is supposed to be affected by four main components, which can be

separated from the observed data. These components are: Trend, Cyclical, Seasonal and

Irregular components. A brief description of these four components is given here. The general

tendency of a time series to increase, decrease or stagnate over a long period of time is termed as
Secular Trend or simply Trend. Thus, it can be said that trend is a long term movement in a time

series. For example, series relating to population growth, number of houses in a city etc. show

upward trend, whereas downward trend can be observed in series relating to mortality rates,

epidemics, etc. while, Seasonal variations in a time series are fluctuations within a year during

the season. The important factors causing seasonal variations are: climate and weather

conditions, customs, traditional habits, etc. For example sales of ice-cream increase in summer,

sales of woolen cloths increase in winter. Seasonal variation is an important factor for

businessmen, shopkeeper and producers for making proper future plans. And The cyclical

variation in a time series describes the medium-term changes in the series, caused by

circumstances, which repeat in cycles. The duration of a cycle extends over longer period of

time, usually two or more years. Most of the economic and financial time series show some kind

of cyclical variation. For example a business cycle consists of four phases, viz. i) Prosperity, ii)

Decline, iii) Depression and iv) Recovery. Irregular or random variations in a time series are

caused by unpredictable influences, which are not regular and also do not repeat in a particular

pattern. These variations are caused by incidences such as war, strike, earthquake, flood,

revolution, etc. There is no defined statistical technique for measuring random fluctuations in a

time series.

1.2 SCOPE OF THE STUDY

The scope of this study will be limited to moving average in time series

using a data on monthly rainfall in Osun State (Oshogbo) from the year 1981

to 2016, using the model selection criteria in selecting the best model for

fitting the plot. Also to test for the stationarity and autocorrelation present in
the time series data used and deselecting the poor model criteria as well as

using the result to forecast for the next 5 years.

1.3 STATEMENT OF PROBLEMS

This study was carried out to check for possible violations of the

underlined assumptions of moving average model and to detect some

limitations of the model.

Konishi & Kitagawa (2008) state, "The majority of the problems in statistical inference can be

considered to be problems related to statistical modeling". Relatedly, Cox (2006) has said, "How

the translation from subject-matter problem to statistical model is done is often the most critical

part of an analysis”. The criteria that will be used in selecting the best model from all the models

available will be our main objective in this project.

1.4 AIM AND OBJECTIVES

This project work is aimed at establishing, comparison and selection of the best Moving

Average model using some model selection criteria with the help of statistical techniques. Thus,

the objectives are to:

i. Reveal the underlining pattern/components of time series presence in the data.

ii. Check whether the selected time series data is stationary or not.

iii. Test for the presence of autocorrelation and adjust if presence

iv. Establish Moving Average models of order q i.e. MA(q).

v. Select the best Moving Average model.

vi. Forecast for the monthly amount of rainfall for Oshogbo, Osun State using the best Moving

Average model.
1.5 SIGNIFICANCE OF THE STUDY

The results of the study will help researchers and the social scientist to

realize the performance of various models in moving average for a time

series data. This study will help to overcome some possible violation of the

underlined assumptions of moving average models. Also, the study will help

the researchers in fitting a model and predict for the future at some point in

time.

1.6 ORGANISATION OF THE STUDY

This study consists of four chapters. Chapter one discussed the Introduction, which

comprises the background of the study, scope of the study, statement of the problem, aim and

objectives and the significance of the study. Chapter two contain the method use in analyzing the

time series data, Chapter three consists of the analysis of the time series data using the method

adopted from Chapter two while Chapter four consist of the conclusion of the whole study.
 CHAPTER TWO

METHODOLOGY

2.1 INTRODUCTION

In this chapter, the methods of analysis to be adopted in performing the analysis of this research

work are discussed. Moving Average method for time series analysis will be discussed and

reviewed.:

2.2 METHOD OF MOVING AVERAGE

Method of moving averages is a very simple and flexible method of measuring trend. It consists

in obtaining a series of moving averages, (arithmetic means), of successive overlapping groups

of the time series. The averaging process smoothens out fluctuations and the ups and downs in

the given data. The moving average is characterized by a constant known as the period or extent

of the moving average.

2.3 TIME PLOT

A Time plot(sometimes called a time series graph) displays values against time. They are similar

to x-y graphs, but while an x-y graph can plot a variety of “x” variables(for example,height,age),

Time plots can only display time on the x-axis. Unlike pie charts and bar charts, these plots do

not have categories. Time plots are good for showing how data changes over time. For example,

this type of charts would work well if you were sampling data at random times.
2.4 AUGMENTED DICKEY-FULLER UNIT ROOT TEST

Time series may be stationary or nonstationary. A stationary time series has the property that the

mean, variance and autocorrelation structure do not change over time.

Augmented dickey-fuller unit root test is one of the tests used to know whether a time series data

is stationary or nonstationary. (That is, if there is need to difference the data to make it stationary

when it nonstationary).

Augmented dickey-fuller t-statistics

Various cases of the test equation:

a. When the time series is flat (that is, doesn’t have a trend) and potentially slow-turning

around zero, use the following equation:

∆ Y t=θ Y t −1−α 1 ∆ Y t−1 +α 2 ∆ Y t −2+ …+α p +at (1)

Where the number of augmenting lags (p) is determined by minimizing the Akaike information

criterion or lags are dropped until the last lag is statistically significant.

b. When the time series is flat and potentially slow-turning around a non-zero value, use the

following equation

∆ Y t=α 0 +θ Y t −1 −α 1 ∆ Y t −1+ α 2 ∆ Y t −2+ …+α p ∆ Y t− p +at (2)

This equation has an intercept term in it but no time trend. Again, the number of augmenting lags

(p) is determine by minimizing the Akaike information criterion or lags are dropped until the last

lag is significant.

The null hypothesis of the Augmented Dickey-Fuller t-test is

H 0 :θ=0 (that is, the data is not stationary)

 Versus

H 1 : θ≠ 0 (that is, the data is stationary)

2.5 MOVING AVERAGE MODEL MA(q)

A moving average (MA) model of order q, denoted by MA(q), is defined by the following

equation

Yt = θ1 e t−1 +...+θq e t−q + et (3)

In the case where only a finite number of the θ-weights are non-zero, we have what we called a

moving average model. In this case,

Y t =e t−θ 1 et −1 −θ2 et −2−…−θq e t−q (4)

We call such a series a moving average of order q and abbreviate the name to MA(q). The

terminology moving average arises from the fact that Y t is obtained by applying the weight 1,

−θ1 ,−θ2 , … ,−θ q to the variables e t , e t−1,e t−2 , … , et −q and then moving the weights and

applying them to e t +1 , e t , et −1,e t−2 , … , et −q+ 1to obtain Yt+1 and so on. Moving average models were

first considered by slutsky (1927) and wold (1938).

Where Yt, t =(1,2,...,T),are the observation values at time t, θ =(θ 1,θ2,...,θq)are arbitrary

coefficients and t ∼ N(0,1) is the white noise innovation at time t. Using the backshift operator

B, on time t, can be expressed as

Y t =θt (B)e t (5)

Where θq(B) = 1 + θ1B + ... + θqBq. If the roots of θq(B) = 0 lie outside the unit circle, then the

MA(q) process is said to be invertible, meaning that yt can be re-expressed as an infinite-order

AR process in terms of θj for j =(1,2,...,q).

The concept of invertibility is exclusively related to the family of MA models and arises

due to the non-uniqueness of the MA autocorrelation functions when θ is replaced by θ−1 , and

can be best described by a simple example using the first order MA model. Consider the

following MA(1) models

A: yt= θet-1 + et

1
B: yt= e + et
θ t −1

It can be shown that these models produce the exact same autocorrelation function (ac.f.), thus

we cannot uniquely identify an MA process from a given ac.f. If we now “invert” models A and

B by expressing et in terms of yt,y t−1,..., we find by successive substitution that

A: et = yt −θyt−1 +θ2yt−2 −...

1 1
B: et = yt − yt−1 + 2 yt−2 −...
θ θ

If |θ| < 1, the series of coefficients of y t−j, j>0, for model A converges, whereas for model B, they

do not. Thus model B cannot be inverted. A general MA(q) model is invertible if the roots of the

equation

θq(B) =1+θ1B +... +θqBq =0

lie outside the unit circle. This is analogous to the stationary restrictions of the coefficients in AR

models.

2.6 AUTOCORRELATION FUNCTION

Autocorrelation are measure of dependence between variables in a time series.

Suppose that Y 1 ,Y 2 , … , Y n are square integrable random variable with the property that the

covariance Yt= Cov(Y t +k , Y t ¿=E [ Y t+ k −E(Y t+ k ) ¿(Y t−E (Y t )) ] of observations with lag k does

not depend on t.

Then,

γk
ρk = , k=0,1, …
γ0

ρk is called autocorrelation function.

Let y 1 , y 2 , … , y n be realizations of a time series Y 1 ,Y 2 , … , Y n. The empirical counterpart of the

auto covariance function is

n−k n
1 1
c ( k )= ∑ ¿ ¿)( y t − y ¿ with y= ∑ y t
n t=1 n t −1

n−k

∑ ( y t +k − y ¿ )( y t − y )
c( k ) t =1
r (k )= = n
¿
c( 0)
∑ ( yt − y ) 2

t −1

The graph of the function r(k), k=0, 1…, n-1 is called correlogram.

2.7 PARTIAL AUTOCORRELATION FUNCTION (PACF)

Definition: Let { X t /t } be a stationary process. The partial autocorrelation at lag h for h ≥ 2 is

defined as the direct correlation between X t ∧ X t−h with the linear dependence between the

Intermediate variables X swith t-h ¿s¿t removed.

Partial autocorrelations measure the degree of association between various lags when the effects

of other lags are removed. If the autocorrelation between Y t and Yt-1 is significant, then we will
also see a similar significant autocorrelation between Y t and Yt-2, as they are just one period

apart. Since both Yt-1 and Yt-2 are both correlated with Yt-1, they are also correlated with each

other; so, by removing the effect of Yt-1, we can measure the true correlation between Yt and Yt-2.

A partial autocorrelation coefficient of order k, which is denoted by α k , is determined by

regressing the current time series value by its lagged values

Y t =b 0+ b1 Y t−1 +b2 Y t−2 +…+ bk Y t−k (6)

2.8 DURBIN WATSON TEST

The Durbin Watson test is a measure of autocorrelation (also called serial correlation) in

residuals from regression analysis. Autocorrelation is the similarity of a time series over

successive time intervals. It can lead to underestimates of the standard error and can cause you to

think predictors are significant when they are not. The Durbin Watson test looks for a specific

type of serial correlation,

The Hypotheses for the Durbin Watson test are:

H0 = no first order autocorrelation. vs H1 = first order correlation exists.

(For a first order correlation, the lag is one time unit).

Assumptions are:

1. That the errors are normally distributed with a mean of 0.

2. The errors are stationary.

Because most regression problems involving time series data exhibit positive autocorrelation, the

hypotheses usually considered in the Durbin-Watson test are

H0 : ρ = 0 vs H1 : ρ > 0

The test statistic is

 n
d=∑ ¿ ¿ ¿
i=2

where ei = yi − ŷi and yi and ŷi are, respectively, the observed and predicted values of the response

variable for individual i. d becomes smaller as the serial correlations increase. Upper and lower

critical values, dU and dL have been tabulated for different values of k (the number of explanatory

variables) and n.

If d < dL reject H0 : ρ = 0

If d > dU do not reject H0 : ρ = 0

If dL < d < dU test is inconclusive.

2.9 CRITERIA FOR MODEL SELECTION

Model selection is the task of selecting a statistical model from a set of candidate models, given

data. In the simplest cases, a pre-existing set of data is considered. However, the task can also

involve the design of experiments such that the data collected is well-suited to the problem of

model selection. Given candidate models of similar predictive or explanatory power, the simplest

model is most likely to be the best choice (Occam's razor). Model selection may also refer to the

problem of selecting a few representative models from a large set of computational models for

the purpose of decision making or optimization under uncertainty.

AKAIKE INFORMATION CRITERION

The Akaike Information criterion (AIC) is a way of selecting a model from a set of models. The

Akaike information criterion (AIC) is a measure of the relative quality of statistical models for a

given set of data. It is defined as:

 AIC=-2Lm + 2m (7)

Where Lm is the maximized log-likelihood and m is the number of parameters in the model.

AICC is AIC with a correction for finite sample sizes. The formula for AIC C depends upon the

statistical model. Assuming that the model is univariate, linear, and has normally distributed

residuals, AICC is defined as:

2 k (k +1)
AICC = AIC+ (8)
n−k −1

Where n denotes the sample size and k denotes the number of parameters.

BAYESIAN INFORMATION CRITERION

The Bayesian information criterion (BICc) or Schwarz criterion is a criterion for model selection

among a finite set of models, the model with the lowest BIC is preferred. It is based, in part, on

the likelihood function. It is defined as

BICC =−2 L m+ 2 log(n)(9)

Where n is the sample size, Lm is the maximized log-likelihood of the model and m is the number

of parameters in the model.

3.0 MODEL VALIDATION

Diagnostic test for the model is carried out to test the adequacy of the selected model. This

involves analyzing possible statistical significance of autocorrelation values of residual. One of

the methods to achieve this is L-JUNG Box Statistics which is described below.

L-JUNG BOX STATISTIC

The L-jung Box statistic, also called the modified Box-pierce statistics is a function of the

accumulated sample autocorrelations, rj, up to any specified time lag m. as a function of m, it is

determined as;

m
rj
Q(m)=n(n+2)∑ ~ X 2 1-α,h (10)
j=1 n− j

Which is approximately X 2m [Link] n is the number of usable data points after any

differencing operations. rj is the estimated autocorrelation of the series at lag m, and m is the

number of lags being tested with h degrees of freedom and significance level α. For significance

level α, the critical region for rejection of the hypothesis of randomness is Q > X 2 1-α,h .

The L-jung Box test is used to test whether or not observations over time are time random and

independent.

H 0 :The autocorrelation up ¿ lag m are all zero

Versus

H 1 : The autocorrelation of one∨more lags differ ¿ zero

 CHAPTER THREE

DATA ANALYSIS AND INTERPRETATION

3.1 Data Presentation

Table 3.1: Annual amount of rainfall (in millimeters) in Oshogbo from 1981 to 2016

Year Annual Rainfall Year Annual Rainfall

1981 1183.5 1999 1607.6
1982 1259.3 2000 1040.5
1983 937.2 2001 1117.6
1984 1865 2002 1683.2
1985 954.7 2003 1421.6
1986 1212.8 2004 1252.9
1987 1796.6 2005 1130.2
1988 1807.8 2006 1469.7
1989 1317.9 2007 1530
1990 1829.9 2008 1597.6
1991 1786.5 2009 1310.2
1992 1403 2010 1692.3
1993 1639.6 2011 1240.5
1994 1043 2012 1191
1995 1605.7 2013 1088.8
1996 1663.7 2014 1649.4
1997 1619.8 2015 1177.1
1998 908.9 2016 1331
Source: Extracted from Nigeria Metrological Agency (NMA) through [Link].

3.2 Time plot

The time plot below shows the pattern of the annual amount of rainfall in Oshogbo, Osun State
Nigeria using the data in the table above.
Annual Amount of rainfall in Millimitres

T imeplot for the annual amount of rainfall in Oshogbo

1800
1600
1400
1200
1000

1980 1985 1990 1995 2000 2005 2010 2015

YEARS

Figure 3.1: Time plot for the annual amount of rainfall (in millimeters) in Oshogbo from 1981 to
2016.

From the above time plot, we can observe that there is no obvious trend and likewise no seasonal
variation as the data is a yearly data. The absence of trend in the time plot is an indication that
the data collected on the annual amount of rainfall (in millimeters) is stationary. To ascertain
this, we will need to perform an Augmented Dickey Fuller test.

3.3 Augmented Dickey Fuller test

Augmented Dickey Fuller Test will be employed to test if the data is stationary or not, and the
hypotheses are given as:

HO: the data are non-stationary Vs H1: the data are stationary

Table 3.2: Result from the Augmented Dickey Fuller test

Augmented Dickey-Fuller Test

Dickey-Fuller Lag order p-value

-3.8208 2 0.03024

Since the p-value is less than the significance value (i.e. 0.03024 < 0.05), the study therefore
rejected the null hypothesis (H 0) and concluded that the data on annual amount of rainfall is
stationary at 5% level of significance as suggested by the time plot above.

3.4 Moving Average model MA(q)

Y t =μ+e t +θ 1 e t−1 +…+θ q et −q

Where Y t is the original series, μis the mean of series, e t is the random error which is assumed to
be white noise, e t−1 is the past random error while θ1 , … ,θ q are the moving average coefficients.
The moving average model is commonly written as MA(q) where q is the order of the moving
average.

3.5 Autocorrelation and Partial Autocorrelation Function plots

The ACF(Autocorrelation Function plot) and PACF (Partial Autocorrelation Function plot) of
the data is shown below:
 1.0
0.8 Autocorrelation Function Plot for the data Partial Autocorrelation Function Plot for the data

0.2
0.6

0.0
0.4

Partial ACF
ACF
0.2

-0.2
0.0
-0.2

-0.4
-0.4

0 5 10 15 2 4 6 8 10 12 14

Lag Lag

Figure 3.2: ACF and PACF plot for the annual amount of rainfall (in Millimetres) in Oshogbo,
Osun State.

From the plots above, most of the spikes are within the confidence limits and this is an indication
that the data is free from autocorrelation. We will need to ascertain this by the use of Durbin
Watson Test.

3.6 Durbin Waston Test

Hypothesis:
HO: there is no autocorrelation Vs H1: there is autocorrelation
 Table 3.3: Autocorrelation test result table for annual amount of rainfall in Oshogbo.

Durbin-Watson test

DW STATISTIC p-value

2.2889 0.7579

The result from the above table shows that the p-value is greater than significant value (i.e.
0.7579 > 0.05), the study therefore do not reject null hypothesis(H 0). And concluded that there is
no serial correlation at 5% level of significance.

3.7 MODEL ESTIMATION

Estimating series of Moving Average models of order q [MA(q)] here, using the annual amount
of rainfall (in Millimeters) and the best model will be chosen by the one with the smallest AIC
and BIC Values. Below is the table containing the estimated Moving Average models with their
AIC and BIC values respectively.

Table 3.4: Model Estimation result

Estimated Models MA(q) AIC BIC

Ŷ t =1400.421−0.1433 et −1 MA(1) 513.38 518.13
Ŷ t =1399.9021−0.1538 et −1+0.0291 e t−2 MA(2) 515.35 521.69

Ŷ t =1397.23−0.1685 et −1+ 0.0504 e t−2+ 0.1182e t−3 MA(3) 516.77 524.68

Ŷ t =1385.8421+0.0659 e t−1−¿ 0.1698e t−2 +0.0261 e t−3 +0,5951 e t−4 MA(4) 517.96 527.46

The study then deduced from the above table that the best Moving Average Model is MA(1)
since it has the smallest AIC (Akaike Information Criterion) and BIC (Bayesian Information
Criterion).
3.8 Model Diagnosis

3.8.1 Histogram for the residuals

Histogram of the model residuals

0.0015
0.0010
Density

0.0005
0.0000

-400 -200 0 200 400

residuals

Figure 3.3: Histogram for the model residuals

The histogram above for the model residuals shows a dumb-belled shape which indicates that the
residuals of the MA(1) model are normally distributed. This is an indication that the model will
be very good for forecasting.
3.8.2 ACF plot for the Model residuals

Series residuals
1.0
0.8
0.6
0.4
ACF

0.2
0.0
-0.4 -0.2

0 5 10 15

Lag

Figure 3.4: Autocorrelation Function Plot for the model residuals

It can be deduced from the plot above that most of the spikes fall within the confidence limits.
This is an indication that the residuals of the model are free from autocorrelation.

Relating to the histogram and the Autocorrelation plots of the model’s residuals above, the study
then concluded that the model is very good and can be used for forecasting.

3.9 Forecast

The forecast below gives the predictions for the next five years with 95% confidence intervals
for the annual amount of rainfall (in Millimeters) in Oshogbo. The MA(1) model gave the
forecast for the year 2017 to 2021.
Table 3.5: Five years Predictions for the annual amount of rainfall (in Millimeters) in Oshogbo,
Osun state.

Year Point Forecast Lower 95% Higher 95%

2017 1201.1 679.7 1722.5
2018 1410.2 887.7 1932.7
2019 1326.8 796.8 1856.8
2020 1425.3 895.2 1955.4
2021 1385.8 771.6 2000.1

Time plot for the forecasted Annual Amount of rainfall

1000 1200 1400 1600 1800 2000
Annual Rainfall

800

1980 1990 2000 2010 2020

Years

Figure 3.5: Forecast plot for the possible amount of annual rainfall in Oshogbo from 2017 to
2021.
From the forecasted annual amount of rainfall, it will rain more in 2020 a little bit more than
what was observed in 2018. The annual amount of rainfall to be observed in Oshogbo this year
(2019) will not be up to that of 2020 and 2021, all with 95% confidence limits.

CHAPTER FOUR

SUMMARY AND CONCLUSION

This project work deals with the comparison and selection of the best Moving Average model.

From the analysis so far, result of the analysis reveals that the annual amount of rainfall in

Oshogbo from 1981 to 2016 was characterized by lack of trend and seasonal variation. Lack of

trend is an indication that the data on the annual amount of rainfall in Oshogbo, Osun state is

stationary. This was ascertained using the Augmented Dickey-Fuller test. Autocorrelation

function plot (ACF) and Partial Autocorrelation function plot (PACF) were plotted and the ACF

plot indicate no presence of autocorrelation in the data. The data was then modeled using

Moving Average Model MA(q). The model MA(1) happens to be the best because it has the

least AIC and BIC values among other models estimated. The model was then diagnosed to see

how appropriate the model suite the data using histogram and the ACF plot for the residuals.

These diagnostic measures showed that the model fits the data and it can be used to forecast for

the future rainfall in Oshogbo, Osun State. Using the established model, five years forecast was

made to know the annual amount of rainfall in Oshogbo Osun State from 2017 to 2021.

Refrences
T. Raicharoen, C. Lursinsap, P. Sanguanbhoki, “Application of critical support vector machine to
time series prediction”, Circuits and Systems, 2003. ISCAS ’[Link] of the 2003
International Symposium on Volume 5, 25-28 May, 2003, pages: V-741-V-744.
 G.P. Zhang, “A neural network ensemble method with jittered training data for time series
forecasting”, Information Sciences 177 (2007), pages: 5329–5346.

Enders, Walter (2004). "Stationary Time-Series Models". Applied Econometric Time Series
(Second ed.). New York: Wiley. pp. 48–107.

Statistical Analysis, Ya-lun Chou, Holt International, 1975, ISBN 0-03-089422-0, section 17.9.

Spencer's 15-Point Moving Average — from Wolfram MathWorld

John H. Cochrane, “Time Series for Macroeconomics and Finance”, Graduate School of
Business, University of Chicago, spring 1997.

K.W. Hipel, A.I. McLeod, “Time Series Modelling of Water Resources and Environmental
Systems”, Amsterdam, Elsevier 1994.

J. D. Hamilton. Time Series Analysis. Princeton University Press, 1994.

J.D. Cryer. Time Series Analysis. PWS-KENT Publishing Company. Boston, 1986

Konishi & Kitagawa (2008, p. 75) and Cox (2006, p. 197)

Montgomery, D. C., Johnson, L. A., and Gardiner, J. S. [1990]. Forecasting and Timeseries

Box, G.E.P., and Jenkins, G.M (1976). Time series analysis, forecasting and control ([Link]), San
Francisco, Holden-Day.