AR estimation in EViews
Posted: Thu Nov 01, 2012 2:06 am
Hello,
I am running a forecasting exercise and part of it involves testing the causality of some different variables (predictors) on a dependent Y. For this reason i estimate a benchmark AR(4) regression of the following form using Newey West standard errors with 12 lags:
eq1.ls(cov="hac",covbw=12) Y c Y(-1) Y(-2) Y(-3) Y(-4) X(-1)
where i test whether the exogenous X enters with a significant coefficient. I thought that running the regression above is the same as estimating: eq1.ls(cov="hac",covbw=12) Y c ar(1) ar(2) ar(3) ar(4) X(-1)
but the t-stat for variable X from the 2nd regression look completely different and therefore the conclusion of predictability. I am convinced that for my purpose the first equation is the most appropriate, but what is really the difference out of curiosity? Thanks in advance.
I am running a forecasting exercise and part of it involves testing the causality of some different variables (predictors) on a dependent Y. For this reason i estimate a benchmark AR(4) regression of the following form using Newey West standard errors with 12 lags:
eq1.ls(cov="hac",covbw=12) Y c Y(-1) Y(-2) Y(-3) Y(-4) X(-1)
where i test whether the exogenous X enters with a significant coefficient. I thought that running the regression above is the same as estimating: eq1.ls(cov="hac",covbw=12) Y c ar(1) ar(2) ar(3) ar(4) X(-1)
but the t-stat for variable X from the 2nd regression look completely different and therefore the conclusion of predictability. I am convinced that for my purpose the first equation is the most appropriate, but what is really the difference out of curiosity? Thanks in advance.