School of Construction Economics and Managements
BUQS2009A
ECONOMETRICS GROUP ANALYSIS
ASSIGNMENT
GR 6
ZAKARIYAA ABDULLAH-2457374
AAAQIL PATEL-2456889
DATASET 1
QUESTION 1
Respone Variable- GDP
a. Multiple modes exist. The smallest value is shown
-The data is not normally distributed or moderately positively
skewed
-0.5<Skewness<1
-mean>mode
QUESTION 2
PRE-TRANSFORMATION
HISTOGRAM
-The Histogram is positively skewed. This is due to majority if the
results filling upon the lower end of the graph.
-Box and Whisker Plot
-The Box and Whisker plot is positively skewed as the median line is closer to the bottom of
the data and the lower whisker is shorter as well.
P-P Plot
-The P-P Plot is not normally distributed as the plots
differentiate from the theoretical straight line
POST TRANSFORMATION
HISTOGRAM
The post transformation histogram portrays normally distributed data as the results
are seen to show symmetrical distribution
Box and Whisker Plot
-The post transformation data shows a concentration above the median and a
dispersion below the median
P-P Plot
The post transformation plots experience less of a deviance than the pre
transformation plots, this results is a more normal distribution
Question 3
To satisfy the linear regression assumptions, the linear regression was performed on
altered data. R-Squared shows that the regression model has 92.9% of the observed
variability in GDP. The level of significance of the GDPI is 31% which shows it
wouldn’t have a major impact on GDP. Government expenditure 3.2% would have a
significant effect on GDP. The Money stock’s level of significance of 66.3% would
also have a significant impact on the GDP. The Treasury Bill having a level of
significance less than 1% is shown as M2 having a larger effect on GDP
Our computed F value from the ANOVA table is 121.671. The corresponding
tabulated F value is 2.63 when the regression degrees of freedom are set to 4, the
residual degrees of freedom are set to 37, and the level of significance is set to 5%.
As a result, we reject the null hypothesis because Fcalc > Ftab: the independent
variables have a large impact on the response variable, GDP.
Question 4
Yes. Because the absence of M2 changes the coefficients of GDPI and TBill, this is
referred to as omitted variable bias. M2 is also worth including since it generates a modest
rise in Adjusted R Squared, implying that the model explains more variation in GDP after
adjusting for degrees of freedom.
Question 5
Analysis of the VIF and Tolerance values reveals multicollinearity. A VIF value larger than 5
implies significant collinearity, and because three of our variables in our model (GDPI,
Government, and M2) all have VIF values more than 5, we can conclude that our model has
multicollinearity. Furthermore, tolerance reveals how tolerable a variable is to collinearity.
Similarly, because the same 3 variables have tolerance levels much below 0.20, we may
corroborate the model's conclusion of multicollinearity.
Autocorrelation
We may test for autocorrelation using the Durbin-Watson test, which indicates that
autocorrelation exists if the test value is less than 2. We do not have autocorrelation when the
test value is bigger than 2. Our Durbin-Watson test value in our model is 0.246, which is far
below 2, indicating that we have autocorrelation in our model.
We may conclude that our model does not display evidence of heteroscedasticity
because there is no trend of growing variance in the residual plots as the predicted
values increase. On a residual plot, heteroscedasticity can be seen as a rise in the
vertical range of the plots as the hypothesized values increase.
Correlation is represented by a number between -1 and 1, with a number closer to
-1 indicating negative corelation and a number closer to 1 indicating positive
corelation. In our model, the correlation between TB and GDPI is -0,435, indicating
a low negative connection between the two variables. (A correlation value less
than -0.5 or greater than 0.5 indicates a high correlation.)
d. No. Because TB is the sole variable that is unaffected by multicollinearity due to
its low VIF value and high tolerance, eliminating it would not be a solution to
multicollinearity. Furthermore, removing TB could cause a much more serious
problem for our model: exclusion of an important variable/model misspecification.
Furthermore, depending on the purpose of the model, we may not need to address
the multicollinearity issue. For example, if the primary purpose of the model is
prediction, we may not need to fix multicollinearity because it has little effect on
prediction or precision.
Question 6
- One can perform principal component analysis by transforming the associated
regressors into orthogonal variables known as principle components. Because these
new components have no corelation, we will use them as regressors in the model. One
may apply partial least square regression to minimize the number of predictors to a
limited set of unrelated new components.One can apply weighted least squares
regression, which involves assigning constants to the observations. This would be
useful because WLS does not always assume linearity in the model.
Question 7
Question 8
The null hypothesis states G has no statistical significance on GDP, such that the
regression coefficient of G is equal to zero.
The determined t value is given by the regression coefficient of G minus the hypothesised
regression coefficient, divided by the coefficient's standard error.
To determine the tabulated t value, we must first determine the level of significance and the
degrees of freedom.We employ a level of significance of 5% for a statistically significant
variable.The number of observations minus the number of parameters gives the degree of
freedom for the residual sum of squares:
With a significance level of 5% and degrees of freedom of 37, the value of t is : 2.0262.
Therefore, since tcalc > ttab, we reject the null hypothesis: G is statistically significant to
GDP.
Question 9
- GDP changes by -0.049 percent on average for every unit change in GDPI.GDP varies by
0.2 percent on average for every unit change in government spending. GDP changes by
-0.017 percent on average for every unit change in M2 money stock.GDP varies by 7.5
percent on average for every unit change in the 6-month Treasury bill rate.
Question 10
lnGDP = 6.370 – 0.00049GDPI + 0.002G - 0.00017M2 + 0.075TB
Question 11
lnGDP = 6.370 – 0.00049GDPI + 0.002G - 0.00017M2 + 0.075TB
lnGDP = 6.370 – 0.00049(250) + 0.002(415) - 0.00017(1360) + 0.075(6.67) = 7.34655
