GPS World: Expert Advice -- Does Anybody Really Know What Accuracy Is?

Expert Advice -- Does Anybody Really Know What Accuracy Is?

http://sidt.gpsworld.com/gpssidt/Expert+Advice+%26+Leadership+Talks/Expert-Advice-mdash-Does-Anybody-Really-Know-What-/ArticleStandard/Article/detail/469052?contextCategoryId=35358

Nov 1, 2007
By:Terry McGurn
GPS World

Does anybody really care? (With apologies to the rock group Chicago.)

As we prepare for the upcoming publication of the 2008 GPS Receiver
Survey in the January issue, I would like to direct everyone's
attention back to the article by Frank van Diggelen in the January
2007 GPS World titled "Lies, Damn Lies, and Statistics." The author
cites the relationships between the various accuracy measures used in
the community and the need to include the metric, as well as the
numerical value when defining accuracy. Ironically, the latest
receiver survey

at that time, appearing in the same issue, demonstrated that many
manufacturers ignore Frank's admonition to include that metric when
specifying an accuracy value.

In my consulting experience, weapons folk usually use circular error
probable (CEP) as the implied measure of accuracy. This leads them to
infer that a number given to them -- lacking a stated metric
-- is a CEP.

I recall when, in the late 1990s, we would discuss the pros and cons
of retaining Selective Availability (SA). Several colleagues in the
Pentagon thought that the 100-meter number agreed to by the
Department of Defense (DoD) and the Federal Aviation Administration
(FAA) was a CEP number -- which is the median or 50 percentile
miss -- when in actuality it was the 95th percentile value of the
measurement scatter, that is, the distance from the aimpoint (or for
surveying, from the survey antenna) containing roughly 95 percent of
the measurements.

So the metric inferred/implied does make a difference. In the late
1990s, while SA was still activated, we took two days of measurements
at a northern Virginia facility. The 95th percentile scatter we
observed was within a meter or so of the 100-meter value. And the CEP
of the demonstrated scatter equated to about 45 meters, generally
consistent with van Diggelen's chart, when you consider the north and
south 1-sigma values of the tested sample were not the same. The
point here is that while SA may have deterred some unauthorized users
from exploiting GPS for landing aircraft, the weaponeers would
probably have killed to get a 45-meter CEP on their weapons or
weapons platforms from GPS.

And that CEP figure was obtained by operating the receiver in a
four-satellite solution mode. As an aside: we continued our tests to
include a receiver operating in an all-in-view mode, similar to the
mode used by non-embargoed civil GPS receivers at the time. When we
ran our receiver in that all-in-view mode the actual median miss
-- that is, the CEP -- of a two-day sample was about 25
meters. So much for SA . . . and the effect of the number of receiver
channels on accuracy.

Before I leave the discussion of van Diggelen's article, I'll take
slight issue with Frank on his assumption that "If you see an
accuracy specified without a metric . . . it is usually CEP." Maybe
that's the case with the military. But, based in part on his own
cell-phone accuracy experience, and my experience in working with the
FAA, that is a very "iffy" assumption.

So, I'm suggesting that we accompany accuracy numbers with the
appropriate metric. But what about presentations to the layman, be he
a congressman who might be trying to determine the relevance of GPS
in cell phones, or the local town board member who's trying to
determine whether it's worth allocating funds to surveying fire
hydrants, or whether all emergency vehicles should be GPS-equipped?
Or a presentation to a defense official who asks whether it's
necessary to have a weapon platform equipped with GPS when the weapon
itself is GPS-equipped?

This audience might well ask the questions in terms of "accuracy."
But what they should be asking is "How does this address my real
problem?" To the question of funding GPS for emergency vehicles, the
proper answer might be: "Sir, what you really want to know is how
fast we will be able to isolate a heart-attack victim in a crowded
parking lot." To the DoD official, it might be: "Sir, given GPS, I'll
tell you the number of weapons we will have to deliver to the target
to ensure a 95 percent probability of destruction."So; help me out
here, Chicago. Does anyone really care about accuracy?

We in the industry sure should, and be precise as to how we define it
-- the first point made here. But I would suggest that we do our
customers a favor and translate that definition into a measure of
effectiveness that the customer is more concerned about, but may not
even know how to articulate. And in going through the process, we
might gain a better appreciation of the effect of accuracy on the
services we're trying to provide.I could go on. Even if the receiver
manufacturers define their metric in GPS World's annual receiver
survey, so many other environmental conditions affect the numbers.
Most importantly, the dynamics.

So, let's be precise amongst ourselves. Shouldn't we educate the
audience to ask the relevant questions, and shouldn't we do our
homework and provide answers to those questions?

Ahh. For the perfect world.

A consultant to U.S. corporations that support the Department of
Defense and the intelligence community, Terry McGurn previously
served for more than 30 years with the Central Intelligence Agency,
most recently as senior analyst for assessing foreign navigation and
guidance technologies. He participated in the drafting of the 1996
Presidential Decision Directive on a national GPS policy, and is
currently a member of the National Space-Based PNT Advisory Board.

Sam Wormley wrote:
Also, if an accuracy value is given, it assumes an excellent
view of the sky. In many situations the view of the
sky is less than "excellent".

--
Dan
(Email: dan at domain below )
(www.gpsmap.net)

The first author's intent appears to be to foster better
communication, especially with a non-technical audience. Although, as
he points out, communications get easily garbled within the
engineering world.

The earliest GPS specs were stated in terms of CEP (I'm reasonably
certain, but don't have those documents available). A few good
reasons -- (1) CEP is the population median, hence is independent of
the eror distribution; (2) In that era, the weapons community used
CEP as an input to JMEM (Joint Munitions Effectiveness Manuals)
methods to calculate weapons effects.

Somewhere along the line, 95th percentiles, 2 drms, etc., replaced the
simpler median.

Besides Van Diggelen's excellent article (glad to see it reprinted),
the Army Corps of Engineers has a nice summary of metrics and
conversions available on-line (free) in Chapter 4 of their GPS
Surveying manual, EM 1110-1-1003. Available at:

http://www.usace.army.mil/publications/eng-manuals/em1110-1-1003/toc.htm

See Table 4-1, including definitions on the page after the basic
table.

On Thu, 08 Nov 2007 18:49:22 -0800, pat_n_ed@yahoo.com wrote:


If you are trying to even "sound" like you know what you are talking
about, you should be discussion SEP as well as or instead of CEP !

wrote:


To answer your question: yes; I do.