DOLS

[rp@limehouse]

I'm a former eviews7.2 user now using eviews8. I was running DOLS regressions on data imported from an excel spreadsheet after it was downloaded from Bloomberg, and one particular price series seems to be messing up any DOLS run with it. When I examined the particular price timeseries in eviews spreadsheet format I detected no errors -- it's monthly data from 1993m01 to 2014m02, and I use the interval from 2001m03 to 2013m12. There were no errors in any of the other timeseries. None of the other DOLS models I run using such downloaded data produces this error (at least none that I've discovered yet). When I go to copy over the DOLS results for equations containing this particular series from eviews spreadsheet view, the regression output (actual, fittted, residuals) skips from 2013m01 to 2013m06. Anything you can think of that would be causing this? I like using excel graphics, which is why I'm copying the eviews 8 output, but it has this skip that renders the output useless. I'm happy to send along a file and direct you to the problem if needs be.

In the years using 7.2 I never had a problem with this. It appears to be only DOLS equations using this particular series (so far anyway) in eviews 8. Do you suggest I go back to 7.2 till this is sorted out?

Oh, p.s.: The copy function does not work in the Estimate dialogue box. Not a big deal, but never a prob in 7.2; and it does slow one down a little.

Thank you.

[EViews Gareth]

Could you email support@eviews.com the files along with a detailed explanation of what is going wrong?

[EViews Glenn]

We just received the file today. Briefly, the user observed that the residual plot for a particular DOLS equation was missing observations for 2013m02-2013m05, and believed this to be an error. After looking at the workfile in question, the behavior is easily explained.

For one particular regressor series, there is a missing value in 2013m03 so that the corresponding difference series has missings in 2013m03 and 2013m04. Since the DOLS specification had 1 lag and 1 lead of each of the differenced regressors, observations for those two observations as well as 2012m02 and 2013m04 are lost, hence the missing residual plot observations.