On Fri, Feb 23, 2007 at 09:19:32AM +0000, Steve Smith wrote:
> Hi Michael,
> This is all correct. The general shape of the filtered model is as
> expected - if you highpass-filter a spike-like event then the
> surrounding values go negative. The variations in the "baseline"
> levels are simply due to the fact that the models are always demeaned
> after filtering etc, i.e. will have zero mean - so, depending on
> what's going on within the model, the "baseline" level may be very
> close to zero or not.
> Don't forget also that the filter is linear, and is applied to both
> the model and the data, so any end-effects, de-meaning effects, etc,
> will do exactly the same to the data as to the model, so all works
> out fine.
> Hope this helps?
It does, thanks.
GPG key: 1024D/3144BE0F Michael Hanke