Test
Results
Decision at 5%
JB test
JB
statistic0.19
Prob. Value
0.91
Errors
normally distributedDW h test
Denominator
0.40
h-statistic
0.64
No autocorrelation
RESET test
Prob. Value (t and F-statistic)
0.46
Model
correctly specifiedARCH test
Prob. Value (F-statistic and Chi-square)
0.86
Homoscedastic variance
As shown in Table 5, the computed JB statistic was found to be 0.19
with a probability value of 0.91. The probability value is higher than 0.05, therefore, the study failed to reject the hypothesised normality assumption and conclude that the residuals are normally distributed.
19
The results from the ARCH test estimated a probability value of 0.86 which is greater than 0.05. Hence the null hypothesis of homoscedastic error variance may not be rejected and conclude there is no problem of heteroskedasticity.
The probability values of the t and the F-statistics from the RESET test were found to be 0.46. Hence, the study failed to reject the null hypothesis and conclude that the whole model is adequate and correctly specified.
Based on these findings, the model seems to be satisfactory in terms of its specification.
As shown in Table 5, the DW h-statistic test estimated a positive denominator of 0.40 and the
DW h
-statistic value was found to be 0.64. Given this DW h
-statistic value, the null hypothesis of no autocorrelation may not be rejected since it is less than the critical value (1.96) and conclude that the model is free from the problem of autocorrelation.
Table 6: Regression Results with Total Output dln(TOUTt) as the dependentVariable
Coefficient
Std. error t-statistic
Prob.
C
1.929820 0.391252 4.932425 0.0000*
d(lnTTOUT
t-1
)
-0.266044 0.138500
-1.92089 0.0658***
lnPT
t-1 0.117526 0.053751 2.186495 0.0380**
lnMP
t-1
-0.355482 0.079114
-4.49330 0.0001*
lnATG
t
0.265710 0.132437 2.006307 0.0553***