Problems about DOLS(Dynamic ordnary least square)

[nikolaichow]

I know that DOLS(Dynamic ordnary least square) model is like normal OLS but additionally included leads and lags of the differenced cointegrating regressors.
However I find that the result is different between I try do the dynamic regression myself and use the eviews option. Why is that happen?
Also since the DOLS option is not available in pool data, could I do the dynamic regression myself by simply included leads and lags of the differenced cointegrating regressors?

[EViews Glenn]

Hard to say about the first since I don't know what you are doing. But there should be a way to make a specification that matches.

As to the second, yes, though you should think about whether you want any deterministics to be heterogeneous or homogeneous.

[wlai4]

Hi,

Im using DOLS at the moment so i was just wondering if i need to truncate the data to allow for the leads and lags in my equation. I have read that in order to account for the leads and lags we need to truncate the data first to allow for the leads and lags in our estimation.

So if i have data from 2000q1 to 2012q4 and i want to estimate a DOLS with 2 lags and 1 lead, do i need to truncate my orignial data to 2000q3 to 2012q3 to account for the 2 lags and 1 lead or does eview automatially adjust the sample.

Also im trying to write a program for the DOLS but it keeps telling me that there is a "near sigular matrix" in my equation. But when i do it using the options it works fine. Can someone please provide advice on what im doing wrong in the program:

smpl est
equation eq_dols_lr.cointreg(method=dols,trend=const,lag=2,lead=0) _
log(y) log(i) log(p) h c s2 s3 s4

Im trying to estimate a DOLS with 2 lags only including a constant in the trend specification. the logs are my dependent varible and the "h c s2 s3 s4" are my deterministic regressors.

thanks in advance for your help.