[daip] SNSMO

[Lynn D. Matthews]

	Amy,

What you say makes sense, and I agree that for a 6 hour smoothing time the
rates should not all have the same values. What worried me specifically
about the 6-hour smooth were the following:

- Last scan for each antenna has a single rate value for all points
  within the scan (and also scan 2 for KP)

- "Slopes" appear across the rate values for a number of scans, while for
   within other scans the points don't have a coherent behavior.

Perhaps my intuition is off because of the effects of the data weights,
but some of these "slopes" don't seem quite right. For example, the fourth
scan of HN slopes up to the right, even though the points within that scan
were originally "random" and the two scans to the right have lower, not
higher mean rates.

Another example: for MK, the first two scans develop slopes in the same
direction (direction of decreasing rates), but a similar trend is not seen
for SC (which also shows a rate decrease with time).

Lynn



On Mon, 13 Feb 2006, Amy Mioduszewski wrote:

> Hi Lynn,
>
> Sorry it took so long to respond but Eric was away last week and I've
> been swamped.
>
> I an looking at your problem.  I think there is a misconception of how
> smoothing works.  Eric can correct me if I am wrong (he'll be back
> Sunday), but I think what happens with boxcar smoothing is each point is
> replaced by the average of the data +-0.5*smoothing time.  So the reason
> that the 20 minute smoothing time replaced everything with a single
> value is that +-10 minutes of each data point is the same data, i.e.,
> your data was spaced in such a way that the average was always based on
> the same points for each.  Now, the 24hr smoothing time replaced
> everything with the same vale for a similar reason, each data point was
> replaced by an average of all the data.  But for a 6hr smoothing time
> each point is going to have a different set of data going into it's
> average, and I think your data is spaced in such a way that this can be
> very different for each point.  SNSMO might still be doing something
> odd, but your suggestion that all the rates should have the same value
> with a 6hr smoothing time is not correct.
>
> Cheers,
>
> Amy
>