Now I have decided to regress log(GFCF) = a+b1R (Nominal interest rate)+b2RR (Real interest rate)+ e(residual)
I have constructed a B-Godfrey test and find that there exists autocorrelation.
The attached pic "removal of autocorrelation" shows how I remove. But my first problem is that the method of first differencing didn't lead me to a significant value (both P-value of D_R and D_RR is too large). Should I still use this regression model to remove the autocorrelation.
Secondly, I used the Augmented Dickey-Fuller test to find if there exist an unit root for all 3 variables, and I found all 3 variables have a unit root of order 1, I(1). As shown in the attached pic "D-F test". My second question is that should I include trend and constant in my D-F test and what does the coefficient of trend actually mean, and does its p-value important (as all 3 trends are not statistically significant).
Afterwards, as all 3 variables have the same order of unit root, I(1). I used the Johansen cointegration test and find that there is a cointegration between them, as shown in the attached pic" Johansencointegration test". My third question is that does it mean that I could use my initial regression model without any changes? This part really confuse me, because I have seen that some previous studies did use the initial regression model while some use the "Error correction model" to regress.
Finally, I tried to use the "Error correction model" and met few problems. As shown in the attached pic "Error correction model", all my dependent variables are not statistically significant and the coefficient of R(nominal interest) is not desired as well (it is expected to be negative). Did I do anything wrong with this model? As it hasn't been taught in my Econometric module and I am not so familiar with it.
The above will be my main body of "Methodology", do I miss anything that is important and do you have any suggestions to my procedures?
Thank you very much for your time
I have constructed a B-Godfrey test and find that there exists autocorrelation.
The attached pic "removal of autocorrelation" shows how I remove. But my first problem is that the method of first differencing didn't lead me to a significant value (both P-value of D_R and D_RR is too large). Should I still use this regression model to remove the autocorrelation.
Secondly, I used the Augmented Dickey-Fuller test to find if there exist an unit root for all 3 variables, and I found all 3 variables have a unit root of order 1, I(1). As shown in the attached pic "D-F test". My second question is that should I include trend and constant in my D-F test and what does the coefficient of trend actually mean, and does its p-value important (as all 3 trends are not statistically significant).
Afterwards, as all 3 variables have the same order of unit root, I(1). I used the Johansen cointegration test and find that there is a cointegration between them, as shown in the attached pic" Johansencointegration test". My third question is that does it mean that I could use my initial regression model without any changes? This part really confuse me, because I have seen that some previous studies did use the initial regression model while some use the "Error correction model" to regress.
Finally, I tried to use the "Error correction model" and met few problems. As shown in the attached pic "Error correction model", all my dependent variables are not statistically significant and the coefficient of R(nominal interest) is not desired as well (it is expected to be negative). Did I do anything wrong with this model? As it hasn't been taught in my Econometric module and I am not so familiar with it.
The above will be my main body of "Methodology", do I miss anything that is important and do you have any suggestions to my procedures?
Thank you very much for your time
Attached Files
Comment