“The likelihood of that happening is about one in 100,” Follow

To expand on what Ugly said: no, you were wrong. Some of us stopped arguing with you because it's obvious that you're invested in your surprising new definition of an established concept.

So yeah, much like Alma and varus.

Werder

You go both ways, eh?

Too late, I have already declared you a gay dude.

I wasn't even paying enough attention to the discussion to know what you're arguing about. I just found it funny that you were pulling a gbaji.

Hating Kachi is so two weeks ago.

I always find it interesting that those statistics are based on an "average number of people killed per passenger mile", or something similar. Which is useful when calculating statistics for an entire population. It doesn't really answer the most relevant question most people are really asking though: "What are my odds of dying if I travel via method A or method B". The absurdity of using passenger miles should be apparent once you realize that the fact that there are more people on a plane doesn't decrease the odds of the plane crashing. Your chance of dying is based on the distance the plane you are in is traveling, and the odds of it crashing over that distance. How many other people may die (or safely make it to their destinations along with you) really doesn't matter.


Silly sidetrack, but I've always found it interesting how such statistics can be manipulated in cases like this.

Yes. The odds of "someone" dying. But not a specific someone.



But not the odds of a single passenger. If I want to know whether *I* am safer driving a car from Los Angeles to New York, or taking a plane, the relevant calculation are the odds of me being a fatality in my car driving that distance, or me being a fatality in a plane flying that distance. The number of other people, while of interest to social statistics on travel modes in general, doesn't actually address the issue of whether it's "safer" from the individuals stand point.

It is incorrect to quote statistics based on passenger miles and make a statement like "Statistics show that you are safer flying in a plane than driving in a car". The reason is because just because one mode may produce a lower death count in total does not mean that it's safer for "you".



There are different odds though. The house can say it statistically stands to make more money on the roulette wheel if more people are betting on it at any given time. But the odds of *you* winning don't change based on the number of people who put their money down on the same numbers you do. You have to understand the effect you're calculating against and how it affects different players differently.



That's not what those statistics are doing though. They are calculating the average number of people who will die while traveling X distance using a given method of transport. Every time a passenger jet carrying 200 people arrives safely at its destination, it counts as though 200 cars carrying a single person driving the same distance had arrived safely. To be fair, every time a passenger jet carrying 200 people crash, it counts as though 200 cars had failed to make the same distance. We could argue that those offset (they wont perfectly), but that's still not the correct way to assess the odds to the individual. The correct method is to calculate the total number of plane crashes over X total miles traveled. This gives you a precise accounting of your odds of being in a plane crash.


Like I said earlier, the use of passenger miles presents us with the absurd assumption that by adding more passengers to a plane, we're reducing the odds that the plane will crash. That's clearly not true.




Oh. that may be true. I'm speaking about the methodology though. It's the wrong way to calculate the individuals odds. It's a useful tool for determining a system wide result, but it tells us nothing about the individual odds of each passenger.



That's not relevant to what I'm talking about though.

Edited, Mar 24th 2011 3:52pm by gbaji

You saying this, doesn't make it so. At the risk of dumbing this down enough for you to grasp what is obvious to some of us:

A man is standing in an airport terminal. He's planning on flying from where he is to somewhere else. He's curious about the odds that he might die in a plane crash. At that moment, for him, his odds are exactly the same as the odds that the plane he's going to get on will crash resulting in his death. Those odds are identical whether he's alone on the plane, or if there are 300 other people on the plane with him. His odds of dying are equal to the odds of the plane crashing in a manner that will result in his death. Period. End of story.

Thus, the odds of any specific traveler dying in a plane crash is equal to the odds of any given plane crashing (fatally) while traveling a number of miles equal to the flight he's taking. If one plane crashes for every 100,000 miles in which planes fly in the US, and he's traveling 1,000 miles by plane, then his odds of dying are 1 in 100.


Counting up the total number of people who've died in plane crashes over a period of time, and then calculating the total number of miles each passenger has flown over that period of time and then dividing one by the other is useful for calculating broad travel statistics, but it does not actually tell the individual traveler his odds of dying on any given trip.


You seriously can't see why? That's just bizarre.


EDIT: Oh. And I think I got the relationship backwards. The calculation you're doing would make his odds of dying increase the more people there are on the plane, not decrease as I said earlier. Here's why:

Passenger miles is a calculation of the total number of passengers who've traveled X distance. Thus, if a plane travels 1000 miles with 1 passenger on it, that's 1000 passenger miles. If that same plane travels the same distance with 300 passengers on it, that's 300,000 passenger miles. If we calculate the rate of fatalities from air travel at say one death per 30 million passenger miles (how many passengers die over time compared to the total number of passenger miles over that same period), then this presents us with a quandary. The plane in the first instance is only going 1000 passenger miles, the plane with 300 people on it is traveling 300,000 passenger miles. Thus, the second case is 300 times more likely to result in fatality based on the statistics (since we're "traveling" 300 times as far relative to the value we're using to calculate our fatality rate). This would lead someone to conclude that he's 300 times more likely to die on that plane than the other.


This is obviously not correct. His odds of dying are the same in each case. What changes is that the "weight" of a plane crash with 300 passengers is greater, since 300 people would die if it crashes. Thus, from a "total air fatality" statistical perspective that plane is at greater risk. But from the point of view of a single passenger, he's not.


Get it yet?

Edited, Mar 24th 2011 7:06pm by gbaji

I'm sure professional statisticians know how to present different aspects of data and the difference between what is essentially an expected value and a probability of an event occurring.

Your example is also a pretty bad one because that's just the marginal probability and not a more accurate conditional probability that takes other factors into account, which really has nothing to do with the difference between what gbaji is talking about.