Hi all,
Im currently working on a strongley balanced panel data set. I have 28 countries and 764 observations. I am looking at what factors influence the level of CO2 emissions in selected countries. All my variables are taken in log form, as i am working with the STIRPAT model. My dependent variable is level of CO2 emissions. My independent variables are as follows, GDP per capita, total population, petroelum prodcts usage of the transport sector and the total urban population. I had first used the fixed effects model to estimate my coeffcients but due to the high levels of multicolinearity i decided to use the ridge regression model.
Using the ridgereg command I was able to use the ridge regression model. Preliminary tests found autocorrelation to be an issue, I used the first difference method to remove autocorrelation. The first difference corrected the issue. However, provided results with very low Rsquared. My original Rsquared provided an Rsquared of 90% and the first difference corrections provides 17.25%.
Output of "ridgereg dLogCO2 dLogGDPperCapita dLogTotalPop dLogPetrolTransport dLogUrbanPop, model(orr) kr(0.1) "
Im currently working on a strongley balanced panel data set. I have 28 countries and 764 observations. I am looking at what factors influence the level of CO2 emissions in selected countries. All my variables are taken in log form, as i am working with the STIRPAT model. My dependent variable is level of CO2 emissions. My independent variables are as follows, GDP per capita, total population, petroelum prodcts usage of the transport sector and the total urban population. I had first used the fixed effects model to estimate my coeffcients but due to the high levels of multicolinearity i decided to use the ridge regression model.
Using the ridgereg command I was able to use the ridge regression model. Preliminary tests found autocorrelation to be an issue, I used the first difference method to remove autocorrelation. The first difference corrected the issue. However, provided results with very low Rsquared. My original Rsquared provided an Rsquared of 90% and the first difference corrections provides 17.25%.
Output of "ridgereg dLogCO2 dLogGDPperCapita dLogTotalPop dLogPetrolTransport dLogUrbanPop, model(orr) kr(0.1) "
Comment