Ahmadi has done a good job by referring you to this document and by reading mine and that, we hope you will be cool.
Let me see if i can make it much more simple.
Let us assume you have 2 variables Y and X. Let's say logged them so lny and lnX. Take take first log or first differences so that we Dlny , Dlnx and let's have a constant (c);
1. Regress lny on c lnx
2. collect the residuals (ect)
3 Dy c ect(-1) Dxt Dxt-1 Dy(-1) Dy(-2) ** Here "t-value of ect(-1) is test of cointegration but you have to include the lags of Dy just enough to avoid autocorrelation.
In #(1) you have the unbiased long-run estimated elasticities but the t-values are incorrect; the short-run elasticities are given by the coefficients on Dx and Dx(-1). The speed of adjustment is given by coefficient on ect(-1).
The first step is to apply the EG/AEG test procedure.In the second step, the first difference of y is regressed on the lagged level of the first-step residual and the lagged first differences of x_1, ..., x_k using OLS. The coefficient on the lagged residual is an estimate of the ECM "speed of correction" parameter. The EG two-step ECM estimation is obtained by specifying the ecm option. Lags of y and x can be included by specifying the lags(#) option: lags(1) hence causing the lagged first difference of y to be added to the second-step ECM alongside the lagged first difference of x; lags(tau) where tau>1 causes lags 1..tau of the first differences of y and x to appear in the second-step ECM.
You should use Stata's regress to estimate the test regression and ECM. All the main regress results are preserved after egranger has run. This allows you to employ Stata's built-in postestimation commands for regress after using egranger. In particular, the standard regress post estimation commands can be used to obtain information criteria and to test for serial correlation in the EG/AEG test regression or ECM estimation regression.
What John Okey Onoh is describing appears tedious to me so let me expand on my answer above.
If Srikanth's aim is simply to test for coint. and stop then with my step above he can simply do this after #2
create the first difference of ect as Dect=ect-ect(-1)
Then as usually run Dec ect(-1) Dec(t-1) Dec(t-2). The significance of ect(-1 ) or the t-value of ect(-1) is a test of E & G cointegration.
In the above step, I went to the error-correction also. Note here there is no constant . I could show the derivation of the critical value using MacKinnon's tables is needed.
Engel Granger cointegration test is for single equation right? It's carried out when the series are of uniform order of integration other than I(0). it is preferable for I(1) series. It's is a single equation variant of Johansen cointegration for systems of equations.
If you are using Eviews;
Conduct OLS on the model and perform unit root on the residual series. if the residual series is I(0), it implies cointegration exist.
replace the residual series with ECM(-1) as ECM = resid using the genr tab.
respecify ur model including ECM(-1) series and estimate. The previous estimates becomes the long-run and the later the short-run with ECM as the speed of adjustment.
This approach is also called Engel-Granger two-step
you offered nothing new, except asking yourself a question. You can talk about how the t-value of the slope can be calculated or why spurious regression? or even how to calculate the critical values
@ Srikanth Potharla, thank you for acknowledging me even though I may have offered nothing new. Prof Chuck A Arize, the teacher of Robert Engel and Clive Granger may as well context it. His name says it all, after all.