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

When you're fundamentally talking about two different things climbing up on your high horse and acting like a pedantic prick serves nothing but your ego and the negative light we all see you in.

Both of you.

4 out of 5 doctors agree with these posts.

I'm quite sure that the professional statisticians understand the difference. But those people usually aren't the ones who write the articles and statements that are presented to the general public. My whole point was about how those statistics are incorrectly presented for public consumption. So statistics using passenger miles will be presented by saying something like "YOU are X times more likely to die traveling in a car than in a plane", when that's not even remotely what the statistics actually show. The statistics actually say that X times more "people" will die per Y miles traveled in a car than in a plane (which is clearly not the same thing). Even ignoring specific variations (being a good driver in a safer car, etc), the statistics those statements are based on aren't the correct ones to use (assuming the intention is to convince someone of the safety of one choice versus another).


That's all I was saying. And it's an accurate statement. I'm not trying to change the freaking world here or anything, just maybe raising awareness about how statistics are often less than accurately presented to the public after they travel through the lenses of media and politics. I've made a similar point in the past about the very real statistical differences between the use of "median" and "mean", but that the two calculations are often used interchangeable in publications meant for general consumption.

Edited, Mar 25th 2011 5:12pm by gbaji

Woah, Gbaji, are you gonna take that? Another rape-related allegation?

You need to stand up for yourselef, man, and the way to do it is legal action.

This.

Uh, Gbaji's not rich in the slightest. You know that, right?

You're rubber, then, are you? I suppose that means Sweetums is glue.

You've regressed.

Yup. That was pretty much all I was pointing out. There's usually a number of different ways to present the same data, and it's not surprising that groups will tend to present things in the best way possible based on what they're trying to sell. It's really based on what you're trying to get and what is important to you.

There's actually an interesting corollary to my own field (lots of fields in fact). We tend to measure failure rates as "device/time". Often this is labeled as MTBF (mean time between failures). That's a measurement of how long a typical device will operate before it croaks and is a relatively useful measurement by itself. But, if you're running a business, what you really want to know is how much you can do with that device before it fails. A hell of a lot of the EDA tools world is about maximizing the amount of work you can accomplish with a given cost of equipment. And that's going to include things like failure rates. Interestingly enough (but not surprising at all), is that the more efficiently you utilizes a piece of hardware, the more work you can get out of it before it fails.

From a cost perspective then, you can calculate "jobs/work per failure" (or some equivalent) in order to measure this. Um... That's an awful lot like the passenger miles per crash statistic above (arguably, it's the exact same thing). What the airline industry is really calculating is how many passengers they can move how many miles before a crash happens. And since profits are based on numbers of tickets sold, they are directly calculating their rate of "loss" from a crash to their rate of "gain" from ticket sales. I would assume they also plan out the size and passenger load of planes to maximize that same transport efficiency, just as we do when planning out the mix of sizes of different computers to run different jobs. You want to maximize the amount you get done over time, while minimizing the relative loss when a system crashes (cause whatever work you're doing on that system is lost and has to be done over, or in the case of a plane crash, you have to pay lots of money to the families of the passengers).

Obviously, human lives aren't the same as a computer design simulation, but the statistical principles involved are pretty much identical.

Yeah, and I'm sorry you've never taken a class in statistics.

Hint: n_1* Pr(x_1) + n_2*Pr(x_2) = average numbers of passengers dying per journey. Pr(x) being probability that the plane crashes on a specific journey, x_1 and x_2 being the events where the plane crashes and doesn't crash, respectively. You can extend this to a probability function relating to miles if you so wish. The main reason I'm not doing this is because I have no ******* idea how to type out an integral on these boards and the miles function would be continuous whereas the journey one is discrete so a sum is sufficient.

n_2*Pr(x_2) is effectively zero if we want to talk about fatalities actually related to the crashing and not Grandpa's heart attack.

n_1 is the number of passengers dying in your average plane crash. That can likely be represented by a probability function as well, but for the sake of brevity, let's leave it out, because it's like, turtles all the way down and sh*t.

Now, I am more interested in the value Pr(x_1), because I'd rather not have to hurtle down to earth, regardless of the fact that I'm not going to die.


PROTIP: I never said deaths by journey was the one passengers wanted, just the one insurance companies used. I said accidents per flight hours would be the one passengers would likely prefer, since it just gives the probability of the plane crashing without consideration for the number of passengers involved.

It was supposed to illustrate that, you know, different statistics for different situations, and that even the professionals will use a wide array of statistics which you seem to be discounting, but you seem stuck on one statistic, so whatever.


Edited, Mar 29th 2011 3:53am by Sweetums

Excuse me if the board doesn't have LaTeX markup like every real board should

I see this thread has degenerated into a correspondence course on statistical analysis...

Statistically, it's much safer.

see what I did thar?