Premium On Safety - Issue 15, 2014 - 7

LESSoNS LEARNED

PART 3 of 4

Pie charts are one of the worst data presentations ever invented
because people think they're much better at interpreting them
than they are.
last year 100 of 1,000 accidents were due
to mechanical failures, but this year things
improved across the board. Mechanical
accidents dropped 10 percent to 90, while
those from other causes fell 30 percent
(from 900 to 630). Abracadabra-the share
of accidents due to mechanical causes
increased by a quarter, from 10 to 12.5
percent (90 of 720). Quoting that last figure
in isolation presents a 10 percent improvement as a 25 percent decline. In a more
familiar (and realistic) example, the proportion of accidents that prove fatal increases
steadily as the certificate levels of the
pilots in command rise from student to
private to commercial to ATP. That doesn't
make flying with an airline transport pilot

more dangerous than flying with a student.
Instead, advanced training (and experience)
vastly reduces the number of swerves and
bounces and hard landings that put aircraft
into the repair hangar each year without
much hazard to life or limb.
Confound It! Perhaps the most common error in data analysis (and the most
disastrous!) is the failure to account for
statistical confounding: the illusion of a
meaningful difference caused by imbalances in other relevant factors. If undercarriage configuration was the only thing
you looked at, you might conclude that
accidents in taildraggers are more survivable than those in tricycle-gear airplanes:
the proportion of fatal accidents is small-

er. But the taildraggers in that accident
record are almost all fixed-gear piston
singles with low cruise speeds. Retracts
and multis, with the higher speeds and
greater impact forces that make accidents deadlier, usually have nosewheels.
Furthermore, taildraggers are more
susceptible to landing accidents, where
serious injuries are almost unknown. A
comparison of non-landing accidents in
fixed-gear singles might make taildraggers look worse-mostly because more
were built without shoulder harnesses.
You get the idea.
Save Pie for Dessert. Pie charts are
one of the worst data presentations ever
invented because people think they're
much better at interpreting them than they
are. Imagine three groups, three categories in each. The largest represents 62,
60, and 65 percent, respectively. Put that
information into side-by-side bar charts
and almost everyone will be able to tell
which is biggest. Display it with pie charts,
and only about one-third will get it right-
what you'd expect to see by pure chance.
Data analysis isn't easy. Worse, the
results usually aren't conclusive. But
knowing some of the ways numbers are
routinely misinterpreted improves your
chance of learning whatever your own
numbers actually have to say.
David Jack Kenny is the Air Safety
Institute's statistician. He is a fixed-wing
ATP with private pilot privileges for helicopters and the owner of a Piper Arrow.

New options in USAIG Performance Vector
USAIG recently welcomed Aircare
International to its Performance Vector
safety initiative. Performance Vector delivers
valuable, forward-looking safety support for
USAIG-insured operators of turbine aircraft.
Eligible policyholders can choose annually
from a range of safety enhancing programs
and services delivered by the industry's
leading providers. The addition of Aircare

International broadens available options to
include aircrew emergency procedures training and mission safety and tele-medical support services.
Aircare International's well-known
Aircare FACTS Training curriculum prepares aircrews for dynamic emergencies
such as cabin fires, medical contingencies, slow onset hypoxia, decompres-

sion, ground evacuation, wet ditching,
underwater egress, and water survival.
Performance Vector-eligible policyholders
can now send pilots to Aircare's OneDay Pilot Emergency Procedures Course,
which uses cabin egress and firefighting
simulators and in-pool drills to cover elements that are especially vital for pilots
who operate without cabin crew, and must
(continued on page 8)
7

Premium On Safety - Issue 15, 2014

Table of Contents for the Digital Edition of Premium On Safety - Issue 15, 2014