(Rich Stowell) wrote
Geez Michael, settle down! So much stress in the cockpit cannot be
conducive to learning or safety...

I don't know about you, but I'm not in the cockpit when I post.

Interesting that I cited a specific source for my statement, which you
summarily ignore as either irrelevant or incapable of leading to
numbers that might be relevant to the concerns that started this post.
Have you read the study I cited?

Yes. It provided no sources on actual engine failure statistics.
Accident statistics are not the same thing at all. Engine failure
statistics from other than GA light piston airplanes also don't cut
it.

If it "hideously skews the picture" wouldn't that apply to all
accident numbers from NTSB?

Yes. NTSB numbers are not a valid way of estimating how often any
event occurs, unless that event always results in an accident or
incident. Actually, my experience with NTSB investigations of light
GA crashes leads me to believe that they're not good for anything at
all.

Each stall/spin accident represent the tip
of the stall/spin problem. Each engine failure accident represents the
tip of the engine failure scenario. Accident stats are a poor measure
of our overall stall/spin awareness, and of our ability to cope with
engine failures precisely because accident numbers represent the
relatively few pilots who have had an accident.

Now there's something we can agree on.

BTW, how do you define relatively few?

But useful information can be gleaned.

Not about the actual rate of incidence of any type of hazard, nor
about relative rates of incidence of various hazards.

Define "rarely." From an industrial accident prevention standpoint,
the theoretical ratio 1:30:300 is often applied wherein for every 331
hazardous encounters of a similar type, only one will progress as far
as an actual accident (significant damage and/or injury). The rest
fall under "incidents" and "hazards."

That ratio is nothing more than an expression of ignorance. In
reality, depnding on the hazard the numbers can be very different.


You neglected to define "rarely."

Consider it equivalent to your definition of "relatively few."

And which "numbers" can be very
different -- total numbers, ratios, what?

In this case, I specifically mean the ratios of accidents and
fatalities to total occurrences. The only numbers we REALLY have are
fatalities - a fatality is difficult to cover up, and thus I would
imagine all (or nearly all - say 98% or better) of fatal accidents are
reported and wind up in the NTSB reports. Non-injury accidents are
often not reported - I know of several where the owner did not have
insurance and did not want to bother with reporting anything. Yes, I
know this is a vilation of NTSB 830. BTW, that includes an engine
failure accident where the airplane was almost completely destroyed.

Granted, total raw numbers
can be significantly different between different accident types,
but--as the study of industrial accident prevention postulates--they
may be linked by comparable ratios or some other normalizing
parameter.

I think the important word here is 'postulate' which of course means
unproven (and maybe unprovable) assertion. Absent proof of a link,
Occam's Razor calls for the least hypothesis - no link.

Industrial safety types love to quote statistics like this to scare
people, but in reality there is usually a reason why some hazardous
encounters lead to accidents or incidents while most do not. It's not
random. These reasons generally have to do with individual skill,
knowledge, and experience as well as factors the industrial safety
people are never told because they involve routine violations of
safety rules. Often the same dynamic plays out in NTSB
investigations.


The intent was not to scare anyone, but to try to add some perspective
tying the comparatively rare accident to the unknown (perhaps
unknowable) number of hazardous situations that are dealt with without
further incident.

And the perspective is flawed.

And yes, there are always reasons why aviation
accidents happen, be it attributable to Software (checklists, SOP's,
etc.), Hardware (airplane, systems, cockpit layout, etc.), Liveware
(the pilot, pax, ATC, etc.), Environment, or the interaction of some
or all of these.

The same is true of industrial accidents. In fact, when you get right
down to it, very few 'accidents' are due to random factors.

In other words, typically 1 out of 331 encounters of a similar type
results in an accident, whether it's precipitated by an engine failure
or an inadvertent stall/spin.

No, this is total nonsense because stall-spins and engine failures are
not similar. First of all, a mechanical failure generally occurs in a
manner that is beyond the pilot's control. When the main seal blows
out, or the engine swallows a valve, or a rod goes through a cylinder,
or the fuel injectors clog with rust - that's almost always completely
independent of pilot skill, knowledge, and judgment. On the other
hand, an inadvertent stall-spin is caused by the pilot. Therefore,
we're not even looking at the same population.

The point of the
pyramidal accident ratio is not to compare engine failures with
stall/spins. Yes, they are two dissimilar accident types in terms of
the driving mechanisms -- the engine in one case vs. the pilot in the
other. But that does not preclude the mix of accidents, hazards, and
incidents within each population from sharing a common relationship.

Well, no - they might share a common relationship through sheer chance
- but that's not the way to bet.

From that standpoint, so what if a hole is blown through the crankcase
and the windscreen gets covered with oil, obscuring the pilot's
ability to see well enough to land under control. The airplane still
gets busted and it's still labelled an engine failure accident.
Likewise, so what if the pilot skids a turn and causes the airplane
to spin into the ground. It's still a stall/spin accident. But for
each one of those accidents, there are many more pilots who, with an
oil-slicked windscreen, were able to land under control; there are
many more pilots who recognized the developing skid, corrected it, and
continued under control. The industrial accident maxim only attempts
to quantify how many within each group were able to avert the
accident.

You can disagree with the theory or its application (in which case, it
would be beneficial to put forth an alternative), but can't you do it
without denigrating?

There's no theory here to disagree with. There is a hypothesis
(advanced without proof) that aviation hazards follow the 1:30:300
distribution, with 331 hazard encounters leading to 30 accidents and 1
fatality (or 30 incidents and 1 accident - depending on how you apply
it). To qualify the hypothesis as a theory, you would need to propose
a logical mechanism for the numerical results. For it to be taken
seriously, you would also need supporting data on the relevant
elements, including some credible data on the rate of hazard
encounters not leading to accidents or incidents. What I am
denigrating here is the attempt to draw conclusions without either.

This is supposed to be a forum for learning -- is
this how you treat your students?

This is certainly how a student in the sciences would expect to be
treated if he tried to pass off the 300:30:1 ratio as a theory.

You're attempting to draw some conclusions about the relative
frequency of stall-spin events relative to engine failure events. No
such conclusions are possible if all you have to look at are accident
statistics.

Just by virtue of the fact that the pilot allowed the inadvertent
stall-spin situation to develop, we can expect that he is less likely
to handle it properly. The same is not true of engine failure.

I would disagree thusly: the pilot who does not routinely ("routinely"
meaning at least 50% of the time) simulate an engine failure followed
by a glide to landing (even from abeam the numbers would be
beneficial) is equally as likely not to be able to handle an engine
failure to a successful landing (i.e.: no accident) as a pilot who
allows the development of an inadvertent stall/spin.

See, this is another example of a hypothesis (I would not even
consider it a theory) that won't stand the light of day. Where did
you come up with 50%? Are you suggesting that a pilot who only makes
20 landings a year (hardly unusual, given how little most private
pilots fly), of which 10 are simulated engine failures, will do better
with a real engine failure than a pilot who makes 300 landings a year,
of which only 30 are simulated engine failures?

It only makes sense that those who don't practice power-off landings
are less likely to be able to competently perform them when necessary,
but going from that to hard numbers without additional evidence is
simply not reasonable.

In the case of NTSB data, one could
extrapolate to get a feel for the order of magnitude of problems
pilots deal with in a particular category by multiplying the number of
accidents by 331.

This is absolutely ridiculous. In addition to the issue of hazard
exposure (mechanical engine failures don't discriminate but
stall-spins do) there is also the issue of hazard magnitude. Off
field landings in gliders, for example, are VERY rarely fatal. The
ratio there is 5000:1. On the other hand, I would be amazed if the
fatality ratio for midairs was much better than 3:1. 331 may be a
good all-around average in aviation (or it may not - data are not
available) but to apply it indiscriminately to all types of hazards
makes no sense at all.

Be gentle, you're dealing with an idiot after all

I doubt it.

Please explain
how an engine failure does not discriminate, yet stall/spin accidents
do. The typical stall/spin profile involves a typical pilot on a
typical flight -- sounds pretty indiscriminate to me.

I don't believe a stall-spin involves a typical pilot at all. I
believe a typical pilot will recognize the loss of airspeed long
before a stall, never mind a spin, actually occurs. The pilot who
allows the situation to deteriorate to the point that an inadvertent
stall occurs is way behind the airplane. Letting it spin is worse.

Of course there are exceptions to this. If you fly just a few knots
over stall in turbulent air long enough (and this is a normal flight
mode for gliders) you will eventually stall. I don't know anyone with
more than 100 hours in a glider who has never stalled in a thermal.
If you fail to control yaw at that point, you will spin, and I know a
few people who have. However, because spin training is still the norm
for glider pilots, and because all glider pilots are aware of the
risk, this situation does not seem to be a significant cause of
fatalities. When glider pilots have fatal stall-spins, they have them
the same way power pilots do - when maneuvering to land.

Please cite your source for the 5000:1 ratio for gliders.

Strictly an estimate based on my experience with the sport. SSA
claims a membership of 14,000. Conservatively speaking, maybe 20% fly
XC, or 2800 people. On average, a XC pilot is going to land out once
a year; more if he's still learning. We have an outlanding-related
fatality once every few years. The last one I recall was Oran Nicks.

Also, glider
pilots are always performing engine-out landings, so it would seem to
make sense that they'd be better at it than those of us who fly
powered airplanes.

Exactly! Further, it's the more skilled glider pilots who go XC and
thus expose themselves to the risk of outlanding. An outlanding in a
glider is an event that discriminates in favor of the more skilled
pilot, and thus I would expect it to have a very low rate of fatality.
On the other hand, a stall-spin discriminates in favor of the less
skilled pilot, and thus I would expect it to have a higher rate of
fatality. Engine failure (due to mechanical problems) does not
discriminate by skill.

As for mid-airs, during the period 1977-1986, 40 percent of the
mid-airs ended without injury.

So my estimate of a minimum 25% fatality rate for midairs (3:1)
doesn't sound too far off. Certainly 1:30:300 is not a good fit.

As for fatality ratio -- yes, the fatality rates between accident
types is not at all equal.

Well, I'm glad we can agree on that. Now if we can simply get to the
next obvious step, that being that the accident/incident rates between
different types of hazards are not at all equal, and that therefore no
conclusions about the rate of occurence of various hazards can be
drawn from accident/incident reports, we'll have reached agreement.

Michael