Cointegration, unit root, stationarity, reliable data?
Posted: Sat Apr 01, 2017 3:42 am
hi, can someone tell me if the results of my equations are valid considering the nonstationary variables, and ADF test for residuals, which says that residuals are stationary.
My variables are:
log_audnzd
log_tot_ratio
log_invest_ratio
bond
log_cpi_ratio
log_audnzd is my dependent variable, all others are explanatory.
When I test for stationarity each variable, they are all non-stationary.
I run the estimation with equation: log_audnzd c log_tot_ratio log_invest_ratio bond log_cpi_ratio
I get this results:
When I want to check the stationarity of the obtained residuals, I generate new variable "REZIDUALI" (GENR, REZIDUALI=resid). I open REZIDUALI file, VIEW, unit root test, level, intercept, SIC, and I get this:
apsolute tau statistics (-3,421228) value is greater than 2,886509, i.e. my residuals are stationary on a 5% siginifacance level, .i.e. the non-stationary explanatory variable I've used are cointegrated? is it a problem if I have in this case 4 EV? is my method right for calculating the model? I will fine-tune (perhaps add more EV later on in order to increase my R/(R2)
Or since I have more than two variables, do I have to perform Johansen methodology for cointegration check? Thank you.
My variables are:
log_audnzd
log_tot_ratio
log_invest_ratio
bond
log_cpi_ratio
log_audnzd is my dependent variable, all others are explanatory.
When I test for stationarity each variable, they are all non-stationary.
I run the estimation with equation: log_audnzd c log_tot_ratio log_invest_ratio bond log_cpi_ratio
I get this results:
When I want to check the stationarity of the obtained residuals, I generate new variable "REZIDUALI" (GENR, REZIDUALI=resid). I open REZIDUALI file, VIEW, unit root test, level, intercept, SIC, and I get this:
apsolute tau statistics (-3,421228) value is greater than 2,886509, i.e. my residuals are stationary on a 5% siginifacance level, .i.e. the non-stationary explanatory variable I've used are cointegrated? is it a problem if I have in this case 4 EV? is my method right for calculating the model? I will fine-tune (perhaps add more EV later on in order to increase my R/(R2)
Or since I have more than two variables, do I have to perform Johansen methodology for cointegration check? Thank you.