Re: A new proof of the superiority of set oriented approaches: numerical/time serie linear interpolation
Date: 3 May 2007 13:45:04 -0700
Message-ID: <1178225104.747192.292320_at_o5g2000hsb.googlegroups.com>
On Apr 29, 4:04 am, Cimode <cim..._at_hotmail.com> wrote:
> I would be grateful for your comments, ideas on this subject....
I've been following this thread with interest over the past few days.
However, it seems there is a glaring problem that I haven't yet seen
It appears that the relation as presented has an overloaded attribute -- namely [ArrivalTime]. This can either be an exact time or an approximate time, depending on whether the value was recorded or interpolated.
As such, it seems the relation should be decomposed such that each of the resulting relations model a single predicate:
Original: "Test unit arrived at train station identified by [ID], which is [Distance] Km from the previous station, at either exactly or approximately [ArrivalTime]."
ID Distance ArrivalTime 24 0.289 2000-01-01 06:29:00.000 25 0.193 2000-01-01 06:29:24.000 26 0.299 2000-01-01 06:29:37.000 27 0.131 2000-01-01 06:29:16.000 28 0.444 2000-01-01 06:29:56.000 29 0.160 2000-01-01 06:29:20.000 30 0.665 2000-01-01 06:30:24.000 31 0.186 2000-01-01 06:33:00.000
(Note that there is no attribute which indicates whether the ArrivalTime is exact or approximate, even though it's stated in the external predicate.)
New1: "Train station is identified by [ID] and is [Distance] Km from the previous station"
ID Distance 24 0.289 25 0.193 26 0.299 27 0.131 28 0.444 29 0.16 30 0.665 31 0.186
New2: "Test unit arrived at station identified by [ID] at [ArrivalTime] time."
ID ArrivalTime
24 2000-01-01 06:29:00.000
31 2000-01-01 06:33:00.000
Report: "Test unit arrived at station identified by [ID], which is [Distance] Km from the previous station, and arrived at [ArrivalTime] which is [ExactOrApprox]."
ID Distance ArrivalTime ExactOrApprox 24 0.289 2000-01-01 06:29:00.000 Exact 25 0.193 2000-01-01 06:29:24.000 Approx 26 0.299 2000-01-01 06:29:37.000 Approx 27 0.131 2000-01-01 06:29:16.000 Approx 28 0.444 2000-01-01 06:29:56.000 Approx 29 0.160 2000-01-01 06:29:20.000 Approx 30 0.665 2000-01-01 06:30:24.000 Approx 31 0.186 2000-01-01 06:33:00.000 Exact
I call this last a "report" because I wouldn't expect it to be updated directly by the user. If, for instance, an additional [ArrivalTime] value was available for one of the "missing" stations, it would be added to the {[ID], [ArrivalTime]} relation.
This design also allows for additional validation of the test procedure. For instance, what if only one sensor worked on the test run? Or what if the only sensors that worked were on adjacent stations -- would it be valid to interpolate beyond the endpoints?
TroyK Received on Thu May 03 2007 - 22:45:04 CEST