Calculating Lags for ECM, HELP please
Posted: Sun Apr 22, 2012 7:42 am
I have performed a 2-step Engle Granger test for cointegration on my 2 variables y on x, both variables are I(1). and i have come out with a long run equation that seems reasonable when described economically.
However i am trying to do an error correction model for the variables in order to come out with the short run equation, but i am constantly being confused by different texts i am reading as to how i should go about performing this.
I first performed this ecm equation: dy dx dy(-1) ecm
where dy - y differenced, dx - x differenced, ecm - residuals of long run lagged by one.
i get an R^2 of 0.13 and Durbin Watson of 2.11
However i read somewhere that to find out number of lags i need to perform an Estimate VAR equation, then look at the Lag Length Criteria, and look at the SC: Schwarz information criterion column, which shows that 2 lags are significant.
Does this mean that my ecm is incorrect and it must be of the form dy dx(0 to -2) dy(-1 to -3) ecm?
i am very confused by this and how do i actually know if the ecm i have is significant or not?
However i am trying to do an error correction model for the variables in order to come out with the short run equation, but i am constantly being confused by different texts i am reading as to how i should go about performing this.
I first performed this ecm equation: dy dx dy(-1) ecm
where dy - y differenced, dx - x differenced, ecm - residuals of long run lagged by one.
i get an R^2 of 0.13 and Durbin Watson of 2.11
However i read somewhere that to find out number of lags i need to perform an Estimate VAR equation, then look at the Lag Length Criteria, and look at the SC: Schwarz information criterion column, which shows that 2 lags are significant.
Does this mean that my ecm is incorrect and it must be of the form dy dx(0 to -2) dy(-1 to -3) ecm?
i am very confused by this and how do i actually know if the ecm i have is significant or not?