we have reached our final model (Figure 3) which only include variables which are significant at 5% level of significance. Thus, data from Figure 3 will be use.

Part 1: Multiple regressionRegression output of Data set ALL After applying backward elimination to Data set ALL regression models Figure 2, we have reached our final model Figure 3 whic

Part 1: Multiple regression

Regression output of Data set ALL

After applying backward elimination to Data set ALL regression models (Figure 2), we have reached our final model (Figure 3) which only include variables which are significant at 5% level of significance Thus, data from Figure 3 will be use

Equation:

Ŷ =b 0+b 1 X 1+ 2 X 2 b

Ŷ = 2,4111 + 0,0003X1 - 0,0407X2 With Ŷ being CO2 emissions (metric ton per capita), X1 being Gross

National Income (GNI) per capita, Atlas method (current US$) and X2 being Renewable electricity output (% of total electricity output)

Interpret the regression coefficient (slope)

b0 = 2,4111 is the intercept coefficient When X1 and X2 = 0, Ŷ= 2,4111 This means if GNI per capita and renewable electricity output is zero, the CO2 emissions will be 2,4111(metric ton per capita)

b1 = 0,0003 and b2 = - 0,0407 are the regression slope coefficient This means that CO2 emissions will increase 0,0003 metric ton when GNI per capita increase by one; and decrease by 0,0407 metric ton when

renewable electricity output increase by one

Interpret the coefficient of determination

R2= 0,6441 is the coefficient of determination This means the variation in the GNI per capita can explain 64,41% of the variation in individuals using the internet rate And the 35.59% remaining is related to other factors outside the research

Regression output of Data set LI

After applying backward elimination to Data set LI regression

models, we have reached a conclusion that all independent values are statistically insignificant This is because even after backward elimination had eliminated down to the last independent variable (GNI per capita), its p-value was still greater than the common alpha level of 0.05, which indicates that this last variable is not statistically significant This claim is further backed-up by data from figure 5, 6 and 7

Regression output of Data set MI

After applying backward elimination to Data set ALL regression models, we have reached our final model (Figure 9) which only include variables which are significant at 5% level of significance Thus, data from Figure 9 will be use

Equation:

Ŷ =b 0+b 1 X 1+ 2 X 2+b 3 X 3 b

Ŷ = 1,5370 + 0,0004X1 - 0,0291X2 + 0,0002X3 With Ŷ being CO2 emissions (metric ton per capita) and X being Gross National Income (GNI) per capita, Atlas method (current US$), X2 being Renewable electricity output (% of total electricity output) and X3 being Air transport, freight (million ton-km)

Interpret the regression coefficient (slope)

b0 = 1,5370 is the intercept coefficient When X1, X2 and X3 = 0, Ŷ= 1,5370 This means if GNI per capita, renewable electricity output and freight air tranport is zero, the CO2 emissions will be 1,5370 (metric ton per capita)

b1 = 0,0004, b2 = - 0,0291 and b3 = 0,0002 are the regression slope coefficient This means that CO2 emissions will increase 0,0004 metric ton when GNI per capita increase by one, decrease 0,0291 metric ton when

renewable electricity output increase by one and increase 0,0002 metric ton when freight air transport increase by one

Interpret the coefficient of determination

R2= 0,5708 is the coefficient of determination This means the variation in the GNI per capita can explain 57,08 % of the variation in individuals using the internet rate And the 42,92% remaining is related to other factors outside the research

Regression output of Data set HI

After applying backward elimination to Data set ALL regression models, we have reached our final model (Figure 12) which only include variables which are significant at 5% level of significance Thus, data from Figure 12 will be use

Equation:

Ŷ =b 0+b 1 X 1+ 2 X 2 b

Ŷ = 3,7700 + 0,0003X1 - 0,1192X2 With Ŷ being CO2 emissions (metric ton per capita), X1 being Gross

National Income (GNI) per capita, Atlas method (current US$) and X2 being Renewable electricity output (% of total electricity output)

Interpret the regression coefficient (slope)

b0 = 3,7700 is the intercept coefficient When X1 and X2 = 0, Ŷ= 3,7700 This means if GNI per capita and renewable electricity output is zero, the CO2 emissions will be 3,7700 (metric ton per capita)

b1 = 0,0003 and b2 = - 0,1192 are the regression slope coefficient This means that CO2 emissions will increase 0,0003 metric ton when GNI per

capita increase by one; and decrease by 0,1192 metric ton when

renewable electricity output increase by one

Interpret the coefficient of determination

R2= 0,4932 is the coefficient of determination This means the variation in the GNI per capita can explain 49,32% of the variation in individuals using the internet rate And the 50,68% remaining is related to other factors outside the research

Part 2: Team regression conclusion

From part 1 data and data from the figures in the appendix, we reached a conclusion that all the models do not have the same significant variables This is because there are no significant variables for low-income countries regression model, and middle-income regression model having one more significant variable (Air transport, freight) than high-income countries regression model and all countries regression model However, there is a similarity, which is regression model from middle, high and all countries all share two similar significant variables, which are GNI per capita and Renewable electricity output

The best regression models for CO2 estimation is the regression model for all countries, since it has the highest coefficient of

determination (R Square) (64,41%) out of the three final regression model, which means the regression model for all countries can estimate CO2 emission with the correct rate of 64,41%

Data set for all countries (ALL)

Data set ALL regression models

Figure 1: Regression models with all variables

Figure 2: Regression models after first backward elimination

Figure 3: Final regression model after second backward elimination

Final Data set for ALL

Data set for Low-Income countries (LI)

Data set LI regression models

Figure 4: Regression models with all variables

Figure 5: Regression models after first backward elimination

Figure 6: Regression models after second backward elimination

Figure 7: Regression models after third backward elimination

Final Data set for LI

Data set for Middle-Income countries (MI)

Data set MI regression models

Figure 8: Regression models with all variables

Figure 9: Final regression model after first backward elimination

Final Data set for MI

Data set for High-Income countries (HI)

Data set HI regression models

Figure 10: Regression models with all variables

Figure 11: Regression model after first backward elimination

Figure 12: Final regression model after second backward elimination