Yes, this is UECM for the F-test., u need one more equation to have ARDL

After estimatio, get b2/b1, b3/b1 and b4/ b1 then create ECM lagged one period

Add to every RHS variables li= 0, not do include that for the lagged dependent variable

the last part of my last comment should read" the lagged dependent variable on the RHS is good so no changes are needed" 

Note that the ECM(-1)  is not lagged residuals. SORRY, I HAVE HAD SO MUCH COMPUTER PROBLEMS THIS MORNING. Feel free to contact me if the above isn't clear to you. Take care

Dear Chuck,

Thanks a lot for your comments. I read them carefully and turn back to you then.

Regards,

Vahid

The specification you provided is the  model for testing cointegration using the bounds approach. The specification is correct for all Variables in the RHS except the lag specification for the short-term dynamics for all the other variables. All the other variables for the short-term part of the expression should have the lag term starting from 0 not 1. You also need two additional models to have ARDL model. thus a (1) long-run model and a  (2) short-run model

Hi, Vahid:  please, go with what I wrote yesterday. Amin's suggestion has been stated in my earlier post,  and he is wrong in line 4 where he writes " All the other  variables .....should have the lag term starting from 0 not 1." If you followed his comments your dependent variable would appear twice, that is, one the LHS and another on the RHS. That will spoil your work and moreover, the computer will not run since you have the dependent variables on two sides of the equation.

I have noted above in quotes that  "the lagged dependent variable on the RHS is good so no changes are needed."  so the lagged dependent variable should start with t minus 1.

 Furthermore, you do not "need two additional models to have ARDL model as Amin stated. I have given you exactly what you need -- From your estimated Unrestricted error-correction model (UECM), you have the first-differenced part as your short-run model or dynamics and the level variables as  your long-run equation, I have shown to you how to  normalize and obtain an ECM term. That is, how it is done. Take care!!

I have attached a sample file to guide you. best of lucks!!!!

Lawrence's attachment is really good, however, you should create the ECM(-1) the way I explained earlier. Get the long-run estimates by normalizing by b1 and multiplying by -1.

My graduate assistants are all gone for summer ( 2 weeks) break so I am not Lawrence, but let us assume that your level terms are  -0.5Ly(-1) + 0.3Lk(-1) + 0.7LF(-1) + 0.3XE(-1) and your constant  -02, Then, create -0.5 * -1 to get .5, then obtain zz = (-02*C + (.3/.5 *Lk)+ (0.7/.5*LF) + (.3/.5 * XE). Your ecm is then  ecm= (Ly-zz) Finally, lag ecm  so you have ECM(-1). Add this term to your first differenced model.

That leads to Lawrence (model 2). Some software like MICROFIT  will automatically do this. The disadvantage is that non-econometricians just use the results and pay no attention to details. 

Dear Lawrence Asamoah Adu,

Your answer helped me in the best way.

Thak you!!

Sorry !! thank you and keep in touch. Chuck

Yes , the form of the model is correct .It is consistent with  Pesaran et al.(2001).