Ethnographic Methodology

CHAPTER THREE
METHHODOLOGY

3.0 Introduction This chapter focuses on the methodology that will be used in the study. The methodology was divided into two parts which is empirical model and econometric methodology. The econometric methodologies that have used include ordinary least square method and diagnostic test. Diagnostic test include stationarity test, multicollinearity test, test for model specification and significance of the model, heteroscedasticity test and correlation test. First, we check for stationary for each variable, if running an OLS regression on non-stationary series results in spurious regression. Second, we will test the variable either multicollinearity or no multicollinearity problem. If the independent variable are

”An alternative but complementary approach to the confidence-interval method of testing statistical hypotheses is the test of significance approach” (Gurajarati& Porter, 2009, p. 115). As studied by Gujarati and Porter (2009), the test statistic and sampling distribution of such a statistic under the null hypothesis is the key idea of t-test. Confidence interval computed as 100 (1-∝) %. Reject null hypothesis (H0) if the test statistic is not in the confidence interval, this means that the test is statistically insignificant. However, not reject the H0 if the test statistic lies in the confidence interval. This means the test is statistically significant.

3.2.2 Test for Stationarity Before running an ordinary least square (OLS) regression, all the variables we used need to test whether the variables are stationary or non-stationary. If running an OLS regression on non-stationary series results in spurious regression. Augmented Dickey Fuller (ADF) test will be used to detect the stationary variables. The hypotheses under the ADF unit root test are:
〖 H〗_0: β=1 (The variable is nonstationary)
H_1: β<1 (The variable is

We reject H_0 in BG test if the test statistic is greater than the critical value. This show that the model is suffers from higher order of correlation. If the test statistic smaller than critical value, H_(0 )is not reject. This means that the model do not have any autocorrelation problem.

Durbin Watson (DW) test

DW test will be used to detect the first order correlation. Time series data mostly involved in regression problem and exhibit positive autocorrelation. The hypothesis usually under Durbin-Watson test is:
H_0 : p = 0
H_(1 ): p > 0
The test statistic is d = (∑_(i=2)^n▒〖(μ_i - μ_(i-1))〗^2 )/(∑_i^n▒μ_i^2 ) or p= (∑_(i=2)^n▒〖μ_i μ_(i-1) 〗)/(∑_i^n▒μ_i^2 ) and, d ≈ 2(1 - p)

Where μ_i = y_i- 〖ŷ〗_i and y_i and 〖ŷ〗_i are, respectively, the observed and predicted values of the response variable for individual i. The serial correlations will increase since d becomes smaller. Upper and lower critical value, d_(u ) and d_L have been tabulated for different of m (the number of independent variables) and n (number of observations).

If d < d_L reject H_0 : p = 0
If d > d_(u ) do not reject H_0 : p = 0
If d_L < d < d_(u ) test is inconclusive or

p = 1 (extreme positive autocorrelation), d ≈