View Single Post
  #121  
Old April 27th 04, 03:44 PM
Dude
external usenet poster
 
Posts: n/a
Default

Dave,

We may have reached the point where we are going to have to disagree on what
we still disagree on.

I believe that picking apart the record is LESS informative than using the
statistic. Here are my reasons:

1. The records of many planes are simply too long to do this with.
2. For any individual accident, there is usually too much mystery to really
assign a cause.
3. The interpretation process is full of opportunities for prejudice.
4. We all think we are above average.

Saying that no one is an average pilot may be true, but I think that the big
stats like this one are useful for all of us. Nothing can predict the
future, but you have to use what data you have to make a decision. I think
that if you take apart the 182 record, you will find a lot of the same
nonsense behavior as in the Cirrus. The value is that less of them died for
it.

After all, we have limitations. Not being Superman, I still need a plane to
fly. Also, I can't spend my life figuring out why Cirrus pilots are fairing
poorly. If I did, I would not be able to have the money to buy one and the
answer would be moot. I am not going to assume that because I have better
judgement, that those things will not happen to me. I am going to exercise
good judgement by buying the plane with a safer record, and then practice
sound judgement again everytime I fly it.


"Dave Katz" wrote in message
...
"Dude" writes:

This stat does paint with a broad brush, but if all you are looking for

is a
measure of average safety in average usage by average pilots (that fly

that
plane) then the measure is very accurate. Yes, if you compare two

models
that are used by vastly different skill levels or in different types of
missions, then you may invalidate the data by means of asking the wrong
question. That is not being done here at all.


But this perverts the nature of statistics. Even very accurate and
valid statistics (with lots of data points, etc.) can never predict
the individual outcome; statistics can only predict the aggregate
outcome. The "average" usage by the "average" pilot does not exist,
and the characteristics of that average cannot even be described, nor
do the statistics predict anything about them. Matter of fact, it is
easy to make the case that the average pilot does *not* scud run in
freezing rain and crash into mountains; it only takes a very small
handful of "special" pilots to skew the statistics.

You cannot even make probabilistic predictions ("I am more likely to
die in a Cirrus than a 182") because the statistics only allow this if
the population is either uniform (like coin flips) or the statistics
can describe the differences between individuals in the population.
The fatalities per 100K statistic is completely worthless for a
quantitative assessment of individual risk. It is meaningful (to some
extent, anyhow) if you are an insurance underwriter, since they deal
in the aggregate, but it only tells them what has been, not what
will be.

Unless you are one of those people who believes you are above average,

then
it means a lot. What makes one person who buys and flies a Cirrus all

that
different from another? What about comparing them to other brands of

new
airplanes buyers? There is no obvious difference, you will have to

propose
one. This is not a picky little nit type of stat. Saying that Cirrus

just
attracts idiot pilots is not enough, you need say why. I haven't seen a
good reason yet.


I don't actually think the Cirrus attracts idiot pilots, that was
someone else's statement. I was trying to use it to make a point.

The statistics (assuming that they pass significance tests) really
tell you only that something is going on, but they can't tell you
what. This is a red flag to go and actually examine the accident
records and try to make an honest evaluation and decide for yourself
what they mean to you.


That would be true, except that examining the records tell us nothing.

You
should rely on the BIG RED FLAG! Seriously. If they had a common

thread
that was fixed, I would grant an exceptional case (aka V tail break

ups).
Until then, no.


Um, the records tell us a lot; the statistic tells us close to nothing
(other than a number.) The records tell us that some of the dead
pilots were scud running in terrible conditions, and if you can
honestly say that you never scud run in terrible conditions, your
personal risk level is much lower than someone who does.

By only looking at the single number, you throw away all of the
information that might help you make an informed risk assessment.

If these planes were mysteriously "falling from the sky" I'd agree
with you, but the failure in most of the cases was squarely in the
left seat.

My point is that the stat is such a large macro that the idiot factor

gets
rounded out. As an average idiot, we are all more likely to die flying

a
Cirrus, than we are flying a 182. We are all average idiots in this

stat.
It is too big to slice apart that way.


See above. Statistically there is no "average idiot" and
mathematically you cannot make the statement that you as an individual
are more likely to die. If the fatality rates were constant, you
could make the statement that more people were going to die next year
per 100K hours in a Cirrus than in a 182, but you couldn't say
anything about your own risk. Furthermore, the fatality rate in the
Cirrus is plummeting as the fleet grows, so if you want to play the
extrapolation game you could predict that the Cirrus rate will be much
lower this year and thus will magically become more safe than the 182.

No, all you have to do is set a standard. How much more risk are you
willing to take on your flight to enjoy the Cirrus over the Cessna? If

its
double, go for it. In my standard, I find the high fatality rate
unacceptable when compared to the ancient Cessna. It should be better.


That's not the same as saying it's either Safe or Unsafe. Life is unsafe,
and you make your risk assessment and live it. The problem is that we are
generally lousy at risk assessment, and ultimately it's somewhat arbitrary
and almost always rationalized (otherwise we'd never get in a car or take
a shower.) I'm probably less safe in an ancient Cessna than in a Cirrus
(this showed the last time I tried to land one!) Is my risk double in the
Cirrus? I doubt it.

I would like the Cirrus fatality rate to be better than it is (and all
other airplane makes, for that matter.) It's hard to judge whether it
"should" be or not. The trends are that it will be, as the number of
fleet hours is growing much faster than linearly (due to the rapid
rate of delivery) but the fatality rate is not keeping pace.

Those are all good, but how does that compare with the Cessna which

requires
a very small time of dual instruction for familiarity? Diamond?

Lancair?
Cirrus gets this level of scrutiny by running around BRAGGING about the

safe
design of their plane with a chute. In his interview I recently read,

Mr. K
was all about how great his airfoil is. Also, they get this scrutiny

because
they have high fatalites.


A new glass panel 182 probably takes almost as much time to transition
into safely for serious use (IFR) as does a Cirrus, though the fact
that so many people learned to fly in them helps with the basic
airwork. The Lancair is in the same class as the Cirrus and I would
expect the transition to be at least as difficult (they're also a bit
faster than the comparable Cirri and the outside visibility isn't as
good.)

Cirrus bragging about safety is rather premature and unfortunate, I agree
with you.

Well, I think the SRV and SR 20 would be better placed in the hands of

more
experienced folk. Since I don't see too many of those folk clamoring

for a
VFR only, glass cockpit, nearly 200k plane, I say they are using it to
attract low time pilots. I think the days of students buying a Cirrus

heve
been nixxed by the insurers.


It's still happening with SR20s; SR22s are pretty much impossible to
insure as a zero-time student, unless you come with a lot of money and
don't expect to solo for a long time.

The SR22 is arguably too much of a handful as a primary trainer,
though a few people have done it. The insurance people are the main
gatekeepers in this case. The number of low-total-time pilots flying
SR22s is probably quite small.


Arguably? Definitely. You may note the 22 is doing better than the 20

in
the stats. I think this is because, as I have heard from more than one

low
time prospective Cirrus buyer, the plane scares them. The 22 must have
higher time pilots at the yoke.


Yep, and a lot of people trade up to the SR22 after building SR20 time.