This Wizard provides dialog-driven access to several cointegration estimation procedures: Johansen maximum likelihood (@JOHMLE), Fully-Modified Least Squares (@FM) and Dynamic OLS (@SWDOLS). A separate wizard is available for using CATS.

Selecting the Cointegration Estimation from the Time Series Menu brings up something like the following dialog box:

Endogenous Variables

Use this field to enter the series you want to test (here the series FTBS3, FTB12 and FCM7). Click on the recursiv.jpg button for a pop-up list of available variables.

Sample Start and End

Use these fields if you want to specify the start and/or end of the sample range. Leave these blank if you want to use the default range.

Estimation Procedure

Select the procedure you want to use from this list (Johansen Likelihood is currently selected).

Deterministic Components

Use this field to select the deterministic terms (if any) included in the model. Choices vary depending on the procedure selected.

Number of Lags

The precise wording of this changes depending upon the procedure. For the Johansen estimator, this selects the number of lags on the differences. For Fully-Modified Least Squares it's the width of the spectral window used in the correction and for Dynamic OLS, it's the number of lags and leads on the differences of the right-side endogenous variables.

Rank of Cointegration Space

For the Johansen and DOLS estimators, it's possible to choose a rank higher than one for the cointegration space. (FMOLS is rank one only). Note that once the rank is higher than one, the "identification" of individual cointegrating vectors requires more than a simple normalization. DOLS creates the cointegrating vectors using a "triangular representation" which uses the first r endogenous variables (in a rank r representation) on the "left side" and the remaining endogenous variables on the right. The Johansen estimator uses an identification based upon construction of the eigenvectors. If you are estimating models which admit more than one cointegrating vector, we would strongly recommend that you use CATS instead, as it allows flexibility in coming up with normalizations with economic content. It also allows for testing overidentifying restrictions.

Error Correction Equations

These allow you to define a VECTOR[EQUATIONS] with the estimated cointegrating relations. This can be used directly on the ECT subcommand for defining a VECM. Note that, as described above in Rank of Cointegration Space, the normalizations chosen are different depending upon the procedure used, so the error correction terms and their loadings in the VECM will be different.