This chapter focuses on the
research method that will be adopted. Regression analysis based on the
classical linear regression model, otherwise known as Ordinary Least Square
(OLS) technique is chosen by the researcher. The researcher’s choice of
technique is based not only by its computational simplicity but also as a
result of its optimal properties such as linearity, unbiasedness, minimum
variance, zero mean value of the random terms, etc (koutsoyiannis 2001,
Gujarati 2004).
MODEL SPECIFICATION
In this study, hypothesis has been
stated with the view of evaluating the effect of government expenditure on
Education in Nigeria.
In capturing the study, these variables were
used as proxy. Thus, the model is
represented in a functional form. It is shown as below:
PL = f (ODA, GR, GDP) ………………………….
3.1
Where
PL = Poverty
level (Dependent variable)
ODA = Office
Development Assistance (Independent
variable)
GR = Grants received (Independent variable)
GDP = Growth
rate of Gross Domestic Product (Independent variable)
In a linear function, it is
represented as follows,
GDP = b0 + b1ODA + b2GR
+ b3GDP + Ut ……………3.2
Where
b0
= Constant term
b1 = Regression
coefficient of ODA
b2
= Regression
coefficient of GR
b3
= Regression coefficient of GDP
Ut = Error
Term
MODEL EVALUATION
At this level of research, using a
time series data, the researcher estimates the model with ordinary least square
method. This method is preferred to others as it is best linear unbiased
estimator, minimum variance, zero mean value of the random terms, etc
(Koutsoyiannis 2001, Gujarati 2004, Baltagi, 1999, and Nwobi 2001).
However, due to conventional reasons,
the researcher will make use of E-view software statistical package in running
the regression. This as believed by the researcher will help in determining the
result of the various tests that is to be carried out. The tests that will be
considered in this study include:
- Coefficient of multiple determination (R2 )
- Standard Error test (S.E)
- T-test
- F-test
- Durbin Watson Statistics
Coefficient
of Multiple Determination
(R2 ): It is used to
measure the proportion of variations in the dependent variable which is
explained by the explanatory variables. The higher the (R2) the
greater the proportion of the variation in the independent variables.
Standard
Error test
(S.E): It is used to test for the
reliability of the coefficient estimates.
Decision Rule
If S.E < 1/2
b1, reject the null hypothesis and conclude that the coefficient
estimate of parameter is statistically significant. Otherwise accept the null
hypothesis.
T-test: It is used to test for the
statistical significance of individual estimated parameter. In this research,
T-test is chosen because the population variance is unknown and the sample size
is less than 30.
Decision Rule
If T-cal > T-tab,
reject the null hypothesis and conclude that the regression coefficient is
statistically significant. Otherwise accept the null hypothesis.
F-test: It is used to test for the joint
influence of the explanatory variables on the dependent variable.
Decision Rule
If F-cal > F-tab,
reject the null hypothesis and conclude that the regression plane is
statistically significant. Otherwise accept the null hypothesis.
Durbin Watson (DW): It is used to test for
the presence of autocorrelation (serial correlation).
Decision Rule
If the computed Durbin
Watson statistics is less than the tabulated value of the lower limit, there is
evidence of positive first order serial correlation. If it is greater than the
upper limit there is no evidence of positive first order serial correlation.
However, if it lies between the lower and upper limit, there is inconclusive
evidence regarding the presence or absence of positive first order serial
correlation.
SOURCES
OF DATA
The data for this research project
is obtained from the following sources:
- Central Bank of Nigeria Statistical
Bulletin for various years.
- Central Bank of Nigeria Annual
Account for various years.
- Central Bank of Nigeria Economic
and Financial Review for various years.