[Lnc-business] Delegate Allocations

[Alicia Mattson]

Looking at the delegate allocations:

1) There appears to be a spreadsheet formula error in the New Hampshire
row, in the "Delegate Allocation % of Members" column.  It ought to be
1.178%, but it is showing 1.200%.  I took a look at your source file, and
the other formulas in the column multiply by 100 before rounding to 3
digits, but that row rounds to three digits before multiplying by 100,
which is why the precision was lost on that row.  This didn't impact the
calculation of their delegate allocation, though, as it wasn't based on
that column.

2)  While the ROUNDUP() function is exactly what should be used for
calculating the "earned delegates" columns (because the bylaws say states
get credit for "fractions thereof"), the percent-of-total calculations will
have better cumulative precision by instead using the ROUND() function.
Unless a division remainder is 0, the ROUNDUP() will always take that final
decimal place to the next higher number, whereas the ROUND() function will
statistically take half of them higher and half of them lower.  As the
results of these roundings are cumulated into a sum, the effect of the
ROUND() function tends to nearly cancel itself out with more and more
numbers in the sum, but the ROUNDUP() function will push the sum higher and
higher as more numbers are added to the sum.

The cumulative effect of the always-round-up approach can be easily
demonstrated.  For the ROUNDUP() function, the sum of all the affiliate
percentages plus the percentage in "Other" comes out to be 100.025%, but
the same sum with the ROUND() function comes out to be 100.001, which is
much more precise.

The difference doesn't impact the way the delegate allocations were
calculated, but if the same ROUNDUP() approach were to be used for
calculating region formation percentages, and a region adds up percentages
to verify that they hit the 10% target, it is more likely that a region
very, very close to the 10% threshold could think that they're over when
they're really slightly under.  It can still happen with ROUND(), but it is
more likely to happen with ROUNDUP().  The regions will avoid the issue
completely by just adding the region's sustaining membership counts and
manually dividing by the total (minus the "other" category), rather than
adding up percentages which have (of necessity) been rounded.  And again,
it's only a risk with regions right on the border of 10% (or multiple
thereof).

-Alicia


On Mon, Dec 2, 2019 at 3:38 AM Caryn Ann Harlos via Lnc-business <
lnc-business at hq.lp.org> wrote:

> Please see attached sent over the weekend to the affiliates.  Regionals,
> please make sure that your state chairs received.  I am not confident of
> the freshness of the state chairs emails available.  It has also been
> posted on the state chairs list.
>
> Mr. Hayes, please make available on the convention site.  It will also be
> available (as well as any subsequent manuals) on lpedia. A Word copy will
> be deposited in the central records repository developed by Mr. Fishman and
> myself.
>
> For future secretaries, I have created a spreadsheet which will
> auto-calculate these items with several "check sum" backcheck features.
> You can view that document here:
>
>
> https://drive.google.com/open?id=1KyisBqMzDFhW-1WzbemyIZRCFwMK-lK7_lHtRGB-hvE
>
> I will be spending a good bit of the time between now and convention
> preparing training materials in the event that there is a new secretary
> elected next convention.
>
> *  In Liberty,*
> * Personal Note:  I have what is commonly known as Asperger's Syndrome
> (part of the autism spectrum).  This can affect inter-personal
> communication skills in both personal and electronic arenas.  If anyone
> found anything offensive or overly off-putting (or some other social faux
> pas) in an actual email, please contact me privately and let me know.  *
>