Estimating error correction model

[linden.luo]

Hi
I am having trouble in estimating an error correction model with a cointegration relationship. Suppose I have y1, y2 and y3 in levels. My equation specification is: dy1 = c(1)*(y1(-1) -c(2)- c(3)*y2(-1)-c(4)*y3(-1))+c(5)*dy1(-1)+c(6)*dy2(-1)+c(7)*dy3(-1). I just applied the ls estimation.
But the estimation results are all wrong. I remembered I used this specification previously and worked fine. OLS is super consistent with cointegration relationships. I wonder why the specification does not work anymore?
I found cointreg command which allows me to estimate a cointegration relationship. Do I have to use this command now to estimate the error correction model above?
Many thanks

[EViews Glenn]

cointreg estimates the model in a different fashion, using the long-run equation with adjustments to handle the cointegrating relationship.

As to what you mean about your equation not working, I'm afraid that I don't understand the issue.