Hardcoding pdl

[denny crane]

Hi,

When I hardcode a PDL specification following the recipe in the Manual (see Special Equation Expressions) to generate a quadratic on a variable and 9 lags (i.e. I write out the equations to match pdl(series,9,2)) and then compare it to the EViews pdl, I can successfully match gamma3 but not gamma1 or gamma2. Strangely, the adj r-square and all other regression stats (including the estimates on non-pdl terms) are the same as if I just used the pdl specification rather than write out the entire lag specification. It seems unlikely that I can follow the recipe for p=2 but not for p=1 or 0 (which is @movsum). Is EViews doing what the Manual says it is doing? Or am I missing something? Thanks.

[EViews Gareth]

Hard to know if you're missing something without being able to see what you've done.

[denny crane]

Sorry... my bad. When moving from Greene's specification (which works for me) to the EViews spec, I mis-calculated c-bar (for want of nail, etc.). Of course, since the gamma3 (squared) term was the result of squaring the (incorrect) gamma2 (linear) term, gamma3 matched the EViews pdl output while the constant and linear terms did not. As the manual clearly states, misrepresenting c-bar ultimately had no impact on the time series model.