Long-run relationship
Posted: Tue Sep 11, 2018 6:48 pm
Hi everybody I am running regression analysis on eviews
I have some questions I wander if you could give me some directions and hints
SO I have panel data (10 countries and 15 years, strongly balanced)
according to Hausman test Fixed effect model appropriate
but I have non-stationary variables (both dependent variable and independent variables)
and all are stationary at first difference.
I run regression analysis with non stationary variables and got quite satisfactory results for myself(significant at level)(output of test is attached)
from one source I found out that even variables are not stationary(assumption: assuming all are stationary at first difference) we could run regression analysis with non-stationary variables and can accept model when our Durbin-Watson statistic is high than our R-squared(in my case it DW statistic is 1.18 and R-squared 0.80) and residuals are stationary(in my case it is stationary). So based on these assumptions can I accept my model as final model and make a conclusion based on test results or should I take another direction. By another direction I meant transform all my variables into first difference and run test again with first differed variables.
Test result was attached.
Please give me suggestion
thank you in advance
R. Aziz
I have some questions I wander if you could give me some directions and hints
SO I have panel data (10 countries and 15 years, strongly balanced)
according to Hausman test Fixed effect model appropriate
but I have non-stationary variables (both dependent variable and independent variables)
and all are stationary at first difference.
I run regression analysis with non stationary variables and got quite satisfactory results for myself(significant at level)(output of test is attached)
from one source I found out that even variables are not stationary(assumption: assuming all are stationary at first difference) we could run regression analysis with non-stationary variables and can accept model when our Durbin-Watson statistic is high than our R-squared(in my case it DW statistic is 1.18 and R-squared 0.80) and residuals are stationary(in my case it is stationary). So based on these assumptions can I accept my model as final model and make a conclusion based on test results or should I take another direction. By another direction I meant transform all my variables into first difference and run test again with first differed variables.
Test result was attached.
Please give me suggestion
thank you in advance
R. Aziz