Questions about Interval VEE rules

Valerie Nibler sent me the following questions regarding interval VEE rules.
I've attached my thoughts to her questions.  Please let me know if you agree or
have a different interpretation by COB Friday, 7/31.  I assume we can
disseminate the clarifications through the UDC/MDMA and MUG groups.....thanks!
(No, the VEE work never finishes!)

Kathy Smith, ABB
919-250-5429
kathy.smith@ustra.mail.abb.com

Questions:
1)  Regarding the statement:  "Intervals containing a power failure
cannot be used as end points for interpolation."  What intervals should
be used as the first and second points for interpolation in the case
where there is a power failure adjacent to the period needing
estimation?

Proposed answer:
My understanding is that an interval with a power failure would be
treated the same way as an interval that isn't valid.  You would find the
closest valid interval that is not a power failure to use as the end point.

2)  Point of clarification on estimating holidays.  It is my
understanding that you should look for holidays in the current and
historical periods, and if you do not find three holidays, you should
look for as many additional Sundays as required to supplement the
holidays.  Average the holidays and the Sundays (a total of three days)
to create the average profile.   A contrasting interpretation is:  look
for holidays first, if you find only one, use that as the "average", but
if you don't find any, then look for Sundays.

Proposed answer:
I think you should always try for three days.  This may end up being a
combination of holidays and Sundays, and current period and historical.  I
think the priority should be:
1.  Holidays in current period
2.  Holidays in historical data
3.  Sundays in current period
4.  Sundays in historical data
If you can't find three holidays or Sundays after all this, then you would
back off to two days or one day, depending on what you can find.

For example, if the current billing period had no holidays, and historical data
had one holiday - use the holiday from historical data, and the two Sundays from
the current billing period that are closest chronologically to the day needing
estimation.  (Assumes all the reference days are "good" data.)





.
Follow-Ups: