How safe is flying REALLY?

It's safe to say that 9/11 has resulted in a dramatic increase in
deaths, due to more people driving instead of flying to their
destination.

A few excerpts from Salon's "Ask the Pilot" series -

"At major airports across America, airplanes come and go at a rate
approaching 100 per hour. Every day in this country, the major
airlines and their affiliates alone operate more than fifteen thousand
flight segments. Of these, almost none fail in their attempt to
successfully defy gravity. During calendar year 2002, not a single
fatality was recorded among the country's commercial airlines - five
million takeoffs and landings by the biggest carriers alone. It's not
always so impressive, but it's always close."

"Earlier this year, Michael Flannagan and Michael Sivak of American
Scientist magazine conducted a study to reevaluate the old
flying-v-driving contention. In the end, their data showed that if a
passenger chooses to drive, rather than fly, the length of a typical
nonstop flight segment (just over 1,100 kilometers), he is now 65
times more likely to be killed."

The last words of Flannagan's and Sivak's report: "For flying to
become as risky as driving, disastrous airline incidents on the scale
of those of September 11th would have to occur about once a month."


Lots of places to review the statistics, among them -

http://www.geocities.com/khlim777_my/ashowsafe1.htm

http://aviation-safety.net/airlinesafety/index.html

Karl Wagner wrote:

Excellent point. I should have thought of that. I did know that all those
people keeping track of numbers at roullete are on a fool's errand ;-)

Yep, roulette tables don't have a memory either. But then again, slot
machines do have a memory. Now there's something to think about...

--
Mikko Peltoniemi
Film & Video Editor, Avid Technician at large.
http://editor.is.dreaming.org

Karl Wagner wrote:

Excellent point. I should have thought of that. I did know that all those
people keeping track of numbers at roullete are on a fool's errand ;-)

Yep, roulette tables don't have a memory either. But then again, slot
machines do have a memory. Now there's something to think about...

--
Mikko Peltoniemi
Film & Video Editor, Avid Technician at large.
http://editor.is.dreaming.org

Mikko Peltoniemi wrote:

Yep, roulette tables don't have a memory either. But then again, slot
machines do have a memory. Now there's something to think about...

It's always entertaining to watch the regulars around slot machines.
They will watch to see if one hasn't paid off in a while, and when the
person operating it leaves, they will pounce, figuring the odds for a
payoff are greater.

In the same vein, there was a story about the Montreal casino. A couple
of the regulars noticed that the cards on a particular machine would
come up in a pattern. Based on that knowledge, they decided to place
some ever-increasing bets to hit it big. They bet twice, and won
something like $30,000, but when they placed the third bet, which would
have won them several hundred thousand dollars, the casino security
detained them for cheating, and confiscated their winnings.

It turned out that the machine relied on a computerized random number
generating routine. This shouldn't have been a problem, as these
generators will truly pick random numbers, except for one minor detail.
Anyone who has done any programming will understand that there are
various ways to start the random number stream, usually by providing a
starting number, called a seed. If the same seed is used each time the
generator is started, exactly the same sequence of "random" numbers will
be produced.

In the case of the machine, it was shut down every night to save power,
and restarted each morning. When the power was turned on each day,
exactly the same seed was used to start the random number generator, so
exactly the same sequence of cards would be played as the day before.
This was how the gamblers were able to predict each hand and win.

As I recall the conclusion, the gamblers were released without any
charges being filed, and a court awarded them their winnings from the
two hands they successfully played. The casino changed the programming
in the machine to make it truly random each day.

Mikko Peltoniemi wrote:

Yep, roulette tables don't have a memory either. But then again, slot
machines do have a memory. Now there's something to think about...

It's always entertaining to watch the regulars around slot machines.
They will watch to see if one hasn't paid off in a while, and when the
person operating it leaves, they will pounce, figuring the odds for a
payoff are greater.

In the same vein, there was a story about the Montreal casino. A couple
of the regulars noticed that the cards on a particular machine would
come up in a pattern. Based on that knowledge, they decided to place
some ever-increasing bets to hit it big. They bet twice, and won
something like $30,000, but when they placed the third bet, which would
have won them several hundred thousand dollars, the casino security
detained them for cheating, and confiscated their winnings.

It turned out that the machine relied on a computerized random number
generating routine. This shouldn't have been a problem, as these
generators will truly pick random numbers, except for one minor detail.
Anyone who has done any programming will understand that there are
various ways to start the random number stream, usually by providing a
starting number, called a seed. If the same seed is used each time the
generator is started, exactly the same sequence of "random" numbers will
be produced.

In the case of the machine, it was shut down every night to save power,
and restarted each morning. When the power was turned on each day,
exactly the same seed was used to start the random number generator, so
exactly the same sequence of cards would be played as the day before.
This was how the gamblers were able to predict each hand and win.

As I recall the conclusion, the gamblers were released without any
charges being filed, and a court awarded them their winnings from the
two hands they successfully played. The casino changed the programming
in the machine to make it truly random each day.

On Sat, 24 Jul 2004 09:01:29 -0700, Reef Fish wrote:

"Karl Wagner" wrote in message ...
Not really b), since the airplane cannot know if and when there has been
a crash lately, or ever for that matter. If there was someone deciding
that Qantas was due for a crash, then it would apply, but since no-one
is, it doesn't apply.

Excellent point. I should have thought of that. I did know that all those
people keeping track of numbers at roullete are on a fool's errand ;-)

Taking it one step further, you should see that the analogy of the
probability of a crash on a flight to the roll of a dice is flawed.

The probability of a particular outcome of a dice-rolling repeated
trial is based on the assumption of a Bernoulli Process check your
statistical dictionary with known, and constant, probability of
"success" on each trial.

A little reflection should convince you that the process governing
the safety of a Qantas flight (or any flight for that matter) is
nothing like that.

No matter what you look at, whenever you use a probabilistic model, you
have no choice but *arbitrarily assuming* the odds.

The real difference is that the *arbitrary* assumption (and knowingly
incorrect, albeit by a small and unknown amount) that all faces of the
dice are equally probable, is close to what it should be. While in the
plane case, we know that there are probably large differences, but we end
up having to assign equal probabilities because we don't know or we have
no way to assess the differences.

So we end up assuming all crashes are equiprobable at least for a given
company, or something like that. We could of course look at the various
companies record, and, say, assign crash odds four times as high to
Singapore Airlines compared with Qantas. And 25 times to China airlines
or Cubana. Maybe two times to AA? Etc.

On Sat, 24 Jul 2004 09:01:29 -0700, Reef Fish wrote:

"Karl Wagner" wrote in message ...
Not really b), since the airplane cannot know if and when there has been
a crash lately, or ever for that matter. If there was someone deciding
that Qantas was due for a crash, then it would apply, but since no-one
is, it doesn't apply.

Excellent point. I should have thought of that. I did know that all those
people keeping track of numbers at roullete are on a fool's errand ;-)

Taking it one step further, you should see that the analogy of the
probability of a crash on a flight to the roll of a dice is flawed.

The probability of a particular outcome of a dice-rolling repeated
trial is based on the assumption of a Bernoulli Process check your
statistical dictionary with known, and constant, probability of
"success" on each trial.

A little reflection should convince you that the process governing
the safety of a Qantas flight (or any flight for that matter) is
nothing like that.

No matter what you look at, whenever you use a probabilistic model, you
have no choice but *arbitrarily assuming* the odds.

The real difference is that the *arbitrary* assumption (and knowingly
incorrect, albeit by a small and unknown amount) that all faces of the
dice are equally probable, is close to what it should be. While in the
plane case, we know that there are probably large differences, but we end
up having to assign equal probabilities because we don't know or we have
no way to assess the differences.

So we end up assuming all crashes are equiprobable at least for a given
company, or something like that. We could of course look at the various
companies record, and, say, assign crash odds four times as high to
Singapore Airlines compared with Qantas. And 25 times to China airlines
or Cubana. Maybe two times to AA? Etc.

You also missed the usual assumption of independence. Probably not a
valid assumption concerning flying.

devil wrote:

On Sat, 24 Jul 2004 09:01:29 -0700, Reef Fish wrote:


"Karl Wagner" wrote in message ...

Not really b), since the airplane cannot know if and when there has been
a crash lately, or ever for that matter. If there was someone deciding
that Qantas was due for a crash, then it would apply, but since no-one
is, it doesn't apply.

Excellent point. I should have thought of that. I did know that all those
people keeping track of numbers at roullete are on a fool's errand ;-)

Taking it one step further, you should see that the analogy of the
probability of a crash on a flight to the roll of a dice is flawed.

The probability of a particular outcome of a dice-rolling repeated
trial is based on the assumption of a Bernoulli Process check your
statistical dictionary with known, and constant, probability of
"success" on each trial.

A little reflection should convince you that the process governing
the safety of a Qantas flight (or any flight for that matter) is
nothing like that.


No matter what you look at, whenever you use a probabilistic model, you
have no choice but *arbitrarily assuming* the odds.

The real difference is that the *arbitrary* assumption (and knowingly
incorrect, albeit by a small and unknown amount) that all faces of the
dice are equally probable, is close to what it should be. While in the
plane case, we know that there are probably large differences, but we end
up having to assign equal probabilities because we don't know or we have
no way to assess the differences.

So we end up assuming all crashes are equiprobable at least for a given
company, or something like that. We could of course look at the various
companies record, and, say, assign crash odds four times as high to
Singapore Airlines compared with Qantas. And 25 times to China airlines
or Cubana. Maybe two times to AA? Etc.

You also missed the usual assumption of independence. Probably not a
valid assumption concerning flying.

devil wrote:

On Sat, 24 Jul 2004 09:01:29 -0700, Reef Fish wrote:


"Karl Wagner" wrote in message ...

Not really b), since the airplane cannot know if and when there has been
a crash lately, or ever for that matter. If there was someone deciding
that Qantas was due for a crash, then it would apply, but since no-one
is, it doesn't apply.

Excellent point. I should have thought of that. I did know that all those
people keeping track of numbers at roullete are on a fool's errand ;-)

Taking it one step further, you should see that the analogy of the
probability of a crash on a flight to the roll of a dice is flawed.

The probability of a particular outcome of a dice-rolling repeated
trial is based on the assumption of a Bernoulli Process check your
statistical dictionary with known, and constant, probability of
"success" on each trial.

A little reflection should convince you that the process governing
the safety of a Qantas flight (or any flight for that matter) is
nothing like that.


No matter what you look at, whenever you use a probabilistic model, you
have no choice but *arbitrarily assuming* the odds.

The real difference is that the *arbitrary* assumption (and knowingly
incorrect, albeit by a small and unknown amount) that all faces of the
dice are equally probable, is close to what it should be. While in the
plane case, we know that there are probably large differences, but we end
up having to assign equal probabilities because we don't know or we have
no way to assess the differences.

So we end up assuming all crashes are equiprobable at least for a given
company, or something like that. We could of course look at the various
companies record, and, say, assign crash odds four times as high to
Singapore Airlines compared with Qantas. And 25 times to China airlines
or Cubana. Maybe two times to AA? Etc.

On Sat, 24 Jul 2004 17:05:55 +0000, Frank F. Matthews wrote:

You also missed the usual assumption of independence. Probably not a
valid assumption concerning flying.

Well, again, the issue is, can we really do better?

I guess this brings us back to the issue of complacency, Alaska Airlines
and their prayer etc. Discussed ealier in the thread.

After a while, one should perhaps get suspicious of airlines which have
been crash-free for a long time. Indeed they might start thinking they
can walk on water.


devil wrote:

On Sat, 24 Jul 2004 09:01:29 -0700, Reef Fish wrote:


"Karl Wagner" wrote in message
...

Not really b), since the airplane cannot know if and when there has
been a crash lately, or ever for that matter. If there was someone
deciding that Qantas was due for a crash, then it would apply, but
since no-one is, it doesn't apply.

Excellent point. I should have thought of that. I did know that all
those people keeping track of numbers at roullete are on a fool's
errand ;-)

Taking it one step further, you should see that the analogy of the
probability of a crash on a flight to the roll of a dice is flawed.

The probability of a particular outcome of a dice-rolling repeated
trial is based on the assumption of a Bernoulli Process check your
statistical dictionary with known, and constant, probability of
"success" on each trial.

A little reflection should convince you that the process governing the
safety of a Qantas flight (or any flight for that matter) is nothing
like that.


No matter what you look at, whenever you use a probabilistic model, you
have no choice but *arbitrarily assuming* the odds.

The real difference is that the *arbitrary* assumption (and knowingly
incorrect, albeit by a small and unknown amount) that all faces of the
dice are equally probable, is close to what it should be. While in the
plane case, we know that there are probably large differences, but we
end up having to assign equal probabilities because we don't know or we
have no way to assess the differences.

So we end up assuming all crashes are equiprobable at least for a given
company, or something like that. We could of course look at the
various companies record, and, say, assign crash odds four times as
high to Singapore Airlines compared with Qantas. And 25 times to China
airlines or Cubana. Maybe two times to AA? Etc.