Comparison between multiple regression and multivariate adaptive regression splines for modelling and forecasting co2 emmissions in Asean countries
Global warming due to the rapid increase in greenhouse gas emissions, mainly carbon dioxide (CO2), is a worldwide issue that leads to escalating pollutions and emerging diseases. This study applies regression analysis to examine interrelationship among the determinants of CO2 emissions. The compa...
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English English |
| Published: |
Universiti Malaysia Sarawak, (UNIMAS)
2015
|
| Subjects: | |
| Online Access: | http://ir.unimas.my/13969/ http://ir.unimas.my/13969/1/Comparison%20between%20multiple%20regression%20and%20multivariate%20adaptive%20regression%20splines%20for%20modelling%20and%20forecasting%20Co2%20emissions%20in%20Asean%20countries%20%2824pgs%29.pdf http://ir.unimas.my/13969/2/Comparison%20between%20multiple%20regression%20and%20multivariate%20adaptive%20regression%20splines%20for%20modelling%20and%20forecasting%20Co2%20emissions%20in%20Asean%20countries%20%28fulltext%29.pdf |
| Summary: | Global warming due to the rapid increase in greenhouse gas emissions, mainly carbon
dioxide (CO2), is a worldwide issue that leads to escalating pollutions and emerging diseases.
This study applies regression analysis to examine interrelationship among the determinants of
CO2 emissions. The comparative performances of multiple regression (MR) and multivariate
adaptive regression splines (MARS) for modelling CO2 emissions in ASEAN countries over
the period of 1980-2007 are discussed. The regression models are fitted individually for every
potential variable investigated so as to find the best-fit parametric or non-parametric
regression model. The results show a significant difference between the performance of MR
and MARS models with the inclusion of interaction terms. The MARS model is
computationally feasible and has better predictive ability than the MR model in predicting
CO2 emissions. Overall, MARS can be viewed as a modification of stepwise regression that
improves the latter’s performance in the regression setting. MARS is better apt to model
situations that involve a large number of variables or a high degree of interaction among the
independent variables. In this study, two scenarios are considered for the forecasting of CO2
emissions for the year 2008-2020. The forecasts of CO2 emissions are computed based on
MARS model using the lagged values of the influential variables obtained from vector
autoregression in the first scenario and the average percentage changes of each influential
variable in the second scenario. The mean absolute percentage error (MAPE) for Scenario 1
and 2 are 1.24% and 3.60% respectively. Scenario 1 has the smaller value of MAPE,
indicating that it reflects the actual data better than Scenario 2. It is projected that there is an
overall increasing trend and the total CO2 emissions, on average, will vary between 51.57 and
62.63 metric tons per capita in the ASEAN Ten based on a 95% confidence interval. The estimated figures show that the amount of CO2 released in ASEAN countries by the year 2020 will be more than double as compared with the 2007 level. |
|---|