Headpigeons!Follow

Eh, considering the leading cause of death for Americans is auto accidents, if I had the choice I'd take my chances.

gbaji, you do know that there are twice as many muggings in California than in New York, right? Not per capita, but actual numbers. Something like to 600k to 300k for 2010. Violent crimes were something like 160k to 80k. Want to explain that?

Explain what? That different areas have different crime rates? I'm still confused as to why you keep bringing this up like a small dog yapping for attention when it has zero relevance to the discussion at hand.


Um... But if we were somehow comparing states rather than urban versus suburban city design, what's your point? California has double the population of New York (so per capita is around the same). Additionally, California has triple the square miles (so per square mile has fewer). So was your intent to show that you're more likely to get mugged in New York than California? Or are you just really really bad at analyzing data?

I'm scratching my head trying to see this "proof". The "real numbers" you provided seem to indicate that California and New York have similar per-capita mugging rates, and New York has about a 50% higher per-square-mile mugging rate. Was that your intention?



Or... you're displaying a remarkable inability to analyze data.

Edited, Jan 18th 2012 2:44pm by gbaji

Oh, look; uninformed analysis of industry metrics. Shocking. If either plane crashed they would have deaths up to their number of passengers minus survivors. Those deaths would be taken into account as part of the 'deaths per passenger mile' statistic. If lets say, there are 20 planes, 10 carrying 20 people and 10 carrying 100 people and each plane flies two 5000 mile trips, of a total 10000 miles in the course of business, it would be a net of 12 million passenger miles. If a 20 cap. plane crashed during the course of business, with a 50% fatality rate, you would end up with 10 deaths per 12 million miles, or 1 death per 1.2 million miles. this is absolutely the correct metric for determining the chance of death, not the chance that any one plane will crash.

If you take the 1/20 crash rate (5%), because 1 out of 20 planes crashed, you'd be massively overestimating your personal chance of death rate, where for a 5000 mile trip it would be like 1/240 (~0.41%).


Edited, Jan 18th 2012 6:51pm by Timelordwho

They're "least significant factors" because they prove you wrong, and "most significant" because they make you feel safer than what reality shows. It's like you have a pull-string on your back, but even Chatty Cathy had more unique lines. I guess if I closed my eyes really really tight and ignored the crime rate, the unemployment, the education rates, the cost of living, and ran face first into a tree and instead of bouncing off just kind of stayed there staring at the tree I could maybe find a reason. Other than that ...

Hah. That's funny. It really is.




Yes. I'm aware of how the statistic "deaths per passenger mile" is generated. My point is that if you forget how the statistic is generated, you can end out trying to do math with it that produces erroneous results. What's funny is that this is exactly what you just did.

That statistic is useful for a broad analysis of the whole industry. If I'm calculating actuarial tables for insurance covering aircraft passengers, I want to know that number. However, it's the wrong statistic if I'm a passenger and I want to know the odds that I will be in a plane crash. This is why I use this example. Because even when I say "this is an example of a statistic that can be misused", someone (like you) will happily oblige us all by insisting on misusing it anyway, thus providing us with an example of such misuse.



No. That's where you are wrong. You are calculating the rate of deaths per passenger mile (1 in 1.2 million), and then dividing by the number of miles you're flying (5000), and saying that you have a 1 in 240 chance of dying. But the actual fact is that you have a 1 in 80 chance of dying.


To derive this, we have to make some assumptions which aren't in the data you started with. Let's assume that each plane has an equal chance of crashing. Let's further assume that in each crash, 50% of the passengers survive and 50% die. In your example, each plane flew 2 equally distant trips (5000 miles). Thus, there were a total of 40 plane trips. Thus, if you are flying on one 5000 mile trip, your odds of being on a plane that crashes during that trip is 1 in 40. Since you have a 1 in 2 chance of dying if you are on a plane which crashes, you thus have a 1 in 80 chance of dying on your trip.


This logic applies in both directions btw. See. If instead of a 20 capacity plane crashing, it had been one of the 100 passenger planes, then the "death per passenger mile" number would have been 5 times higher (since 50 people died instead of 10, again sticking to our 1/2 die assumption). So 1 death per 240k miles. With that set of data, you might conclude that the odds of dying while traveling 5000 miles is therefore 1 in 48. But it's not. It's still 1 in 80. Because the odds of a single specific plane crashing does not change based on the number of people traveling in it (which is what I said at the beginning). A plane will crash or not crash based on other factors. Your odds therefore are based on the odds of those factors happening.


Do you see how the odds of *you* dying in a plane crash isn't the same as the average deaths over distance divided by the distance you're traveling? It's a mathematical calculation which appears on the surface to make perfect sense, but if you step back from the data and think about what you're actually calculating, it's wrong. In this particular example, the passenger miles stat is useful if you are an insurance company and want to know how many payouts you'll have to make for every X passengers flying Y miles so you can correctly charge people. But it's not useful at all to determine individual odds of dying in a plane crash.

Edited, Jan 18th 2012 4:41pm by gbaji

Yes, I can agree with this.

The person in the urban area still has a 1 in 200 chance per year of being mugged, however rarely they go into a dangerous square mile per year, compared to a rural person having a 1 in 100 chance of being mugged, no matter how often or how little they go into a "dangerous"="safe" square mile. Yes the absolute figures are 50 muggings in the urban square mile, that the urban person walks through. Yes the absolute figure is 10 muggings per square mile that the rural person walks through. The urban person and the urban perpetrators are simply diffused through more people than the rural ones, making the absolute numbers higher but leaving the relative numbers of people chances of victim-hood alone.

Some people think that since a poker machine might have a one in million chance of paying out the jackpot, if they stick to one poker machine that hasn't Jackpotted in a long time, then the poker machine's "time" to pay out must be coming up. Or that if they stick to one poler machine their entire life and hit it more than a million times that they MUST get the Jackpot eventually. Statistics don't work like that. Real life statistics of what happen in the real universe don't work the way that human "pattern recognising systems" think/feel that they do. Every time that one poker machine is hit, there is a one in million chance of the jackpot. That poker machine may well not Jackpot in 1 billion years of constant hits. Or that Poker machine may very possibly Jackpot 50 times in a row.

Another example that is counter human intuitive: 1, 2, 3, 4, 5, 6, is a random number string as well as a sequential integer string.

Get it?

Except that you failed to account for the increased likelihood that you will be on one of the high capacity flights that did not crash.

I didn't miscalculate.


Edited, Jan 19th 2012 4:24am by Timelordwho

Well, here are the violent crimes reported in some popular, fairly walkable neighborhoods in Houston (I live here) that have decent transportation compared to the rest of the city, anyway. Keep in mind that these are heavily populated areas, and that Houston doesn't have the best reputation safety-wise (although the statistics are largely influenced by the really, really poor areas, which I'm too lazy to list) so although the incidents may be even an order of magnitude more, this is still far, far less than 1% of the population in these areas. Also remember: even if the crime is on the street, it's not necessarily perpetrated by a stranger!

I used the police department's crime map, which while a bit cumbersome compared to raw data, but it works well enough for me.

The last 30 days:

Midtown

Murder: no incidents returned
Rape: no incidents returned
Aggravated Assault: out of 11 of them, only 4 of them were outside, with 3 on the road/sidewalk and 1 at a gas station parking lot. The rest were in bars and strip clubs. This is my surprised face.
Robbery: Out of 9 incidents, 7 were on the sidewalk, and most of these were really fucking late. As in, past midnight, when basic common sense dictates that you shouldn't walk. The others were robbed convenience stores and either a bar or a club. It's not super-specific.

Montrose

Murder: no incidents returned
Rape: 1, in an apartment. Remember, the vast, vast majority of rapes are perpetrated by someone the person knows, and no change of location will prevent an acquaintance, friend, or lover from proving themselves to be an amoral @#%^.
Aggravated Assault: 5 on the street. However, you have to keep in mind that there are a lot of gay bashings in this neighborhood, since it's basically the gayborhood. For some reason, there are homophobic straight dudes that like to come here to beat the shit out of gays.
Robbery: out of 11 of them, only 4 were on the streets, and all of them were on largely empty side streets. The rest? Convenience stores.

University Place

Murder: no incidents returned
Rape: no incidents returned
Aggravated Assault: Out of 4, only 1 was on the road. At midnight. The rest were at bars (do you still see my surprised face?).
Robbery: 5, and all of them were at places of business

Downtown

Murder: no incidents returned
Rape: no incidents returned
Aggravated Assault: 8, but only 4 of these were on the road or in parking lots.
Robbery: There were 4 of these, 3 of which were outdoors, and 2 of these were late at night in the fucking forest.

Now, do you want to look at the car crash statistics? I'll list the ones for Harris County for the year 2010, provided by the Texas Department of Transportation, to include some of the Houston suburbs, but many also live in other counties so it doesn't paint a complete picture. If you want to complain about it being for the whole year, just divide by 12.

Crash Statistics for Harris County

Fatal Crashes: 335
Fatalities: 361
Serious Injury Crashes: 8,035
Serious Injuries: 11,196
Other Injury Crashes: 14,201
Other Injuries: 23,577
Non-injury Crashes: 42,550
Unknown Severity Crashes: 2,300
Total Crashes: 67,421


Now, just for comparison's sake, here's a quick overview of causes of death from the CDC. Motor vehicle accidents were at #96, murder was around #107.


So not only do more people die from motor vehicle accidents than murders, most murders are not even perpetrated by strangers! So how is not staying out late going to protect you from those you know? Here's a list by gender: (Statistics from the US Department of Justice)

Women
Intimate partner: 30%
Other family: 11.8%
Acquaintance: 21.8%
Stranger: 8.7%
Undetermined: 27.6%

Men
Intimate partner: 5%
Other family: 6.8%
Acquaintance: 35.3%
Stranger: 15.5%
Undetermined: 37.4%

So really, who's actually safer?




Edited, Jan 19th 2012 10:46am by Sweetums

Well, they're not at all the deepest darkest places of Houston, they're actually fairly popular urban neighborhoods.

Edited, Jan 19th 2012 1:16pm by Sweetums

That is not true at all.

I had a feeling you'd misunderstand what I wrote. Thought about editing it, but got distracted.


I meant that your odds "over time" of being mugged are higher if you walk down that dark alley every day of your life. Obviously, your odds that one night are no higher. Given that you want to avoid being mugged, then not doing that very often is a good idea, right? Not doing it ever is even better.


Um... Also, your odds "that night" of being mugged are higher if you walk down the dark alley versus if you stay in your home. Which ties into the point I've been making all along. When you look at crime statistics, you can't assume that this means everyone in the area over that period of time has the same odds of having been a victim of a crime. What you do in that area during that time period matters a hell of a lot. Your mistake through this entire discussion has been that you believe that if X muggings occur out of Y people over Z time, that the odds of *you* being mugged in that area over that time is X/Y.


But that is simply not true. You can't apply those sorts of statistics to the individual condition and outcome. It simply falls apart.

Edited, Jan 19th 2012 5:42pm by gbaji