Interpreting Cointegration, IRF and VDC Analysis
Posted: Thu May 31, 2012 12:13 am
Hello there. I am quite new here. I am using Eviews 6 to analyze a monetary transmission mechanism. My variables are:
1) Industrial Production Index
2) Mudharabah Interbank Investment (a money market instrument)
3) Islamic finance (loan)
4) Overnight Policy Rate (interest rate adopted by the central bank)
5) Islamic interbank money market rate
6) Consumer Price Index (CPI) - exogenous variable
I am not sure if it is wise to put in the CPI variable as exogenous variable. The variable is included based on other past research.
Firstly, for the unit root test, do we need to transform the variable to the natural logarithm first or simply run the unit root test without the variables transformed in the natural logarithm form?
Secondly. I am not sure whether I am using the right procedure for the cointegration test. This is because I was taught in order to select the max lag length is to estimate using the unrestricted VAR and view the lag length criteria based on the AIC. Another method I read told to estimate using the multivariate VECM model by clicking quick - estimate VAR - VECM and estimate the VECM start from lag 1, then check the residuals and autocorrelation of the error term to obtain the lag length. Can anyone please tell me what is the right procedure for the cointegration test. Supposed my cointegration test is correct, how do I interpret the output and what if the sign of the cointegrating coefficients are incorrect?
Thirdly, in order to generate the IRF graphs and also the Variance Decomposition analysis (table), I am not sure whether I should pick Analytic (asymptotic) or Monte Carlo with 1000 repetitions for the Response Standard Error. And for the Impulse Definition - Decomposition method, what is the difference between cholesky (dof adjusted and generalized impulses? This confusion arises because my lecturer taught me not to change any of the settings in Eviews for the IRF and VDC. However, my literature review revealed that a similar past research use Monte Carlo with 1000 repetitions and generalized impulses. In addition, it would be helpful if you could help me interpret my analysis.
Hereby, I attached some of the output (results) of the Johansen cointegration test, Granger causality test, Impulse Response Function graphs and Variance Decomposition analysis table.
Your help is very much needed and I would like to express my gratitude and appreciation in advance. Thank you and have a pleasant day!
1) Industrial Production Index
2) Mudharabah Interbank Investment (a money market instrument)
3) Islamic finance (loan)
4) Overnight Policy Rate (interest rate adopted by the central bank)
5) Islamic interbank money market rate
6) Consumer Price Index (CPI) - exogenous variable
I am not sure if it is wise to put in the CPI variable as exogenous variable. The variable is included based on other past research.
Firstly, for the unit root test, do we need to transform the variable to the natural logarithm first or simply run the unit root test without the variables transformed in the natural logarithm form?
Secondly. I am not sure whether I am using the right procedure for the cointegration test. This is because I was taught in order to select the max lag length is to estimate using the unrestricted VAR and view the lag length criteria based on the AIC. Another method I read told to estimate using the multivariate VECM model by clicking quick - estimate VAR - VECM and estimate the VECM start from lag 1, then check the residuals and autocorrelation of the error term to obtain the lag length. Can anyone please tell me what is the right procedure for the cointegration test. Supposed my cointegration test is correct, how do I interpret the output and what if the sign of the cointegrating coefficients are incorrect?
Thirdly, in order to generate the IRF graphs and also the Variance Decomposition analysis (table), I am not sure whether I should pick Analytic (asymptotic) or Monte Carlo with 1000 repetitions for the Response Standard Error. And for the Impulse Definition - Decomposition method, what is the difference between cholesky (dof adjusted and generalized impulses? This confusion arises because my lecturer taught me not to change any of the settings in Eviews for the IRF and VDC. However, my literature review revealed that a similar past research use Monte Carlo with 1000 repetitions and generalized impulses. In addition, it would be helpful if you could help me interpret my analysis.
Hereby, I attached some of the output (results) of the Johansen cointegration test, Granger causality test, Impulse Response Function graphs and Variance Decomposition analysis table.
Your help is very much needed and I would like to express my gratitude and appreciation in advance. Thank you and have a pleasant day!