Schneier on Security
A blog covering security and security technology.
« More Low-Tech Security Solutions |
| Court Limits on TSA Searches »
July 7, 2009
Why People Don't Understand Risks
Yesterday's Minneapolis Star Tribune had the front-page headline: "Co-sleeping kills about 20 infants each year." (The headline in the web article is different.) The only problem is, in either case, there's no additional information with which to make sense of the statistic.
How many infants don't die each year? How many infants die each year in separate beds? Is the death rate for co-sleepers greater or less than the death rate for separate-bed sleepers? Without this information, it's impossible to know whether this statistic is good or bad.
But the media rarely provides context for the data. The story is in the aftermath of an incident where a baby was accidentally smothered in his sleep.
Oh, and that 20-infants-per-year number is for Minnesota only. No word as to whether the situation is better or worse in other states.
Posted on July 7, 2009 at 1:50 PM
• 73 Comments
To receive these entries once a month by e-mail, sign up for the Crypto-Gram Newsletter.
"The issue was raised afresh in May when, police say, a Lakeville grandmother mixed alcohol and pills, then accidentally slept on her 6-week-old grandson and killed him."
Co-sleeping didn't kill the baby, mixing pills and alcohol killed the baby.
I heard this one from someone who considers themselves a very reliable source: "14% of all passenger aircraft ever built have crashed with loss of all lives." References? Context? Huh?
I read about this a while ago. If you filter out alcohol and drug users, the rates are no worse than any other sleeping arrangement.
It's quite commonplace in other cultures. It's just in North America that we've been brainwashed into thinking it's dangerous.
Full disclosure: we're doing it. Well, actually just Mommy and Junior in the big king bed now. I got banished to a mattress on the floor :-( We do it because the boy won't sleep any other way.
And this article also doesn't seem to differentiate between caretakers who plan for their child to sleep in their bed and caretakers who happen fall asleep (or pass out) with their child in their bed.
Or more accurately: this article doesn't seem to actually be about co-sleeping, but about caretakers who mix drugs, alcohol, and babies, and one possible outcome.
@Piper "It's just in North America that we've been brainwashed into thinking it's dangerous."
I agree. We're do it sometimes, but not when alcohol or cold medicine are on the menu--it's separate beds then. Is that sensible risk reduction?
" If you filter out alcohol and drug users, the rates are no worse than any other sleeping arrangement.
It's quite commonplace in other cultures. It's just in North America that we've been brainwashed into thinking it's dangerous."
Remindes me of the so Called "Fan Death" in South Korea:
"Fan death is a South Korean urban legend which states that an electric fan, if left running overnight in a closed room, can cause the death of those inside (by suffocation, poisoning, or hypothermia)."
@Dan: Thanks for the example, "14% of all passenger aircraft..."
I often argue with people about risk evaluation, and lately (since the Air France crash in the Atlantic) these arguments have been about aviation, a topic that seems to bring out most wonderful irrationality (for reasons that Bruce has often written about in this blog).
The claim from your self-proclaimed reliable source tickled my funny bone. How would one verify such a claim? A Herculean labor, perhaps. As you observed, without references or context, what use could be made of such "information"?
It reminded me of the words of C. F. Gauss, probably the greatest of all mathematicians: "I could easily lay down a multitude of such propositions, which one could neither prove nor disprove"
Interesting mention of "Fan Death" in this context. Recent research led to a recommendation that leaving a fan near a sleeping infant reduces the risk of Sudden Infant Death, by reducing the amount of CO2 that can build up near the baby's face.
"14% of all passenger aircraft ever built have crashed with loss of all lives."
no doubt this was true at some point in the history of aviation.
Its a conspiracy to sell more baby-cribs!
Here is another one to put the airplane thing into perspective: There are more people alive today, than have ever died. Quite possible something like this holds for airplanes as well, maybe with a large margin. (Reasoning of whether you will be reincarnated into this universe or more likely not left as exercise to the reader...)
Incidentially, 14% does not sound too bad, considering that they can do that only once in their often very long career. The one crashing with french passengers was 20 years old.
A similar misuse of statistics occured when we were in a vaccine trail for Lyme disease. I got the placebo, my wife the real shot. She subsequently got Lyme, I didn't. However the newspapers reported that the vaccine was 100% effective for women over 50, when the actual answer was 99.9%. One may not round in this case as the impression of never failing is different than rarely fails.
It's not just drugs/medicins/alchol but also smoking and some other medical conditions (including sleep apnia) that increase the risk.
By and large if you are a healthy non smoker then sleeping with a baby is not to much of a problem (watch out for bed cloth smothering though).
If the mother is feeding the tot the natural way then it is possible with the little one in the bed for her to only partialy wake to feed junior and get better rest. Incidently the tot usually sleeps better as well especially if skin to skin contact with either parent is involved.
Of those that I know who have done this their children apear to be healthier and happier and better behaved than others (and yes this is just a perception not a study with any maths basis).
Well there was one additional statistic:
"But Thompson, who sits on a statewide committee that reviews child deaths, said she recommends that parents never sleep with infants. At least 50 percent of infant deaths in the county are caused by co-sleeping, she said."
Nothing given to back it up, but presumably Ms. Thompson has evidence. At least there's a source given.
The lack of context is, I am certain, quite deliberate. Fear and panic sell newspapers and content. If you put things in perspective, there's no reason for panic, so no sale of newspaper.
I'm glad to see Lyme Disease taken seriously. Last time I checked, the medical community was still bickering over whether it exists at all.
At least the article does give some indication of the comparative risk:
"Statewide, it's the third most common cause of death in babies less than a year old, behind premature births and birth defects."
and "At least 50 percent of infant deaths in the county are caused by co-sleeping, [Thompson] said.
*Chronic* Lyme Disease is still being debated. Sorry.
It's funny the slant we see on news stories in a relatively well-developed country. More than half of the human race still sleeps on dirt floors the way we have for thousands of years, and I bet very few of those parents spend much time worrying about getting a big, fancy crib so their baby doesn't die from co-sleeping.
14% of all passenger aircraft ever built -- how many are that? It's at least 3500, taking into account only the number of DC-3's and Boeing 737's built (because those figures are easy to find). Has there been 35 fatal aircraft crashes a year on average for the last 100 years?
My intuitive answer would be no -- that's almost one crash a week, which would tend not to make crashes headline news -- but http://www.planecrashinfo.com/cause.htm claims to know of 1300 fatal accidents in the period 1950-2008. That's within a (generous) order of magnitude ...
Well, 14% of all passenger aircraft may in fact be correct. Any aircraft more than nine seats, (depending on your source - some list 19 seats as the magic number) qualify as passenger aircrafts. There are non-commercial aircrafts that crash in rural areas or outside the country that we may never read about or see.
"14% of all passenger aircraft ever built have crashed with loss of all lives."
The wording is a little ambiguous - maybe that actually means that 14% of all TYPES of passenger aircraft ever built have crashed with all loss of lives. If you list all the series of passenger aircraft (ie, B737, B747, B757, A330, MD-80, etc, and remember to separately count the different types in a series: B737-300, B737-400, etc) then I'd believe that 14% of those have had an accident in which all passengers/crew had been killed.
This isn't the point of the original post but co-sleeping is quite an issue in NZ and there appears to be some very good data available. I comment mostly to make this information available to those who might possibly need it.
The interesting part is that the Maori providers were very resistant to the banning of bed sharing prefering a solution of a flax basket that could be used on the same bed. An interesting solution to a cultural problem.
A very useful overview article from NZ Listener magazine (a vaguely left wing middle class weekly) http://www.listener.co.nz/issue/3601/features/...
The formal information from the Ministry of Health.
Some interesting links to stats from an interest group.
This stuff bugged the hell out of me when my wife and I were in parenting classes. They would give us statistics that say things like, "if you breastfeed, your baby is less likely to die from SIDS." and I would say, "Wait... is that because breast milk helps prevent SIDS or is it because parents who breastfeed are also parents that are less likely to put their children in SIDS related situations?" Nobody could tell where the statistics came from or why so much inference was where it was. There is so much causal leaping in statistical analysis of this kind that it makes the statistics laughable.
About the 14% of all passenger aircraft...
I could easily believe that 1 in 5 small craft eventually crashes and kills everyone on board, after many hundreds or thousands of flights (rather than being, for example, retired to a junkyard after 30+ years of faithful service).
For commercial airliners crashes are a lot less frequent, but there are probably only a few thousand commercial airliners in the world (?). Somehow I expect there are at least 10x as many small craft as there are jumbo jets etc.
So the 14% figure might possibly be accurate.
Of course, it is a meaningless statistic because the odds that YOU will be in one of those planes when it happens to crash, are infintesimally small. I'm too lazy to look it up, but I would expect to find that the average plane that suffers a fatal crash, had flown hundreds or (more likely) thousands of successful flights between the time it was manufactured, and the time of the crash. A small plane that makes one flight per month for 20 years, will have 240 flights without a fatal crash. If it made one flight per week for 30 years, that's more than a thousand flights. Only the LAST one of which involved the flaming death of all aboard... I mean, its pretty obvious that 14% of all plane *flights* don't end in a fatal crash, isn't it?
"But Thompson, who sits on a statewide committee that reviews child deaths, said she recommends that parents never sleep with infants. At least 50 percent of infant deaths in the county are caused by co-sleeping, she said."
I don't understand how co-sleeping can be a cause. Suffocating by getting entangled in bedding is a cause, or by being smothered by an adult. The question is whether co-sleeping increases the risk of this type of death.
Considering that co-sleeping has occurred since the dawn of humanity, I find the idea that it's risky to be non-intuitive. You might call it "natural". On the other hand, drugs and alcohol are very "unnatural".
Either way, leave it to a statistic to say a lot without actually saying anything.
Well there you go again Bruce, applying critical thinking. How do you expect to be fear mongered into submission (or prosecution) if you go thinking about data sources and all the facts about a hypothesis?
Thanks for the great blog
It's actually penetrated into the medical profession. When you are close to ready to leave, a nurse comes and interviews you, asking how you are going to have your child sleep. The correct answer is in a separate bed on her back. Both of our children have ended up on their stomachs a few weeks into their lives (once we were sure they didn't sleep too deeply and could lift their heads). It's a good thing - they rest better. Both sleep next to their mother after eating in the middle of the night. It's a good thing - they rest better.
@derick: while I'm in agreement with your general point, the argument that something's been done since the dawn of humanity holds little water for me. Dying in infancy has been done since the dawn of humanity. Frequently. The rate at which it happens has gone down dramatically in recent centuries. Eons, millennia, or centuries ago, not co-sleeping was dangerous to infants because of things such as predators and hypothermia before the advent of houses with central heating and cooling. Of course, I'd guess that those took a backseat to disease, and possibly malnutrition, depending on the time and place...
But the point is, once you eliminate the things that have caused the truly frightening rates of infant mortality that humanity has endured throughout most of our history, then it's entirely possible that co-sleeping is the most dangerous thing for an infant, so long as there are reasonably safe alternatives. Or, maybe it still isn't. The other point is that a simple count of deaths is a completely useless statistic for judging whether it is or isn't.
In rem sleep some people exhibit muscle contaction that leaves bruises. Potentially dangerous if hands around a neck. Lesson, keep hands away from all necks when sleeping.
People also sleep walk, a weird thing that is kind of hard to believe unless you have known someone with the condition.
I think sleeping around infants is a dumb thing, cribs exist for a good reason.
Melatonin might be a dangerous thing for people dumb enough to sleep around infants and not understand the rem effects. No warnings on any labels yet.
Risks management is not easy, it takes understanding and processing of sporadic events with many different combinations of lifestyle.
For what it's worth, the Very Reliable Source is a member of a prestigious political body in a large Western European country. He presented the factoid in a casual discussion of modern air travel, and suggested---without actually stating it---that he was referring to large passenger aircraft. This doesn't really change the current discussion, but I thought people might be interested.
"Very Reliable Source" is very unreliable.
How many airliners crash with loss of all life per year? If we're talking 737/A320 or larger, such an event is world news. From my unreliable memory, I'd say that is about 2-4 per year, but for arguments sake, lets call it 12 - one per month. An airliner has a commercial lifetime of 20-30 years - again, to be generous, call it 30 years. The 14% number implies a probability of about 0.0004 per month of crashing with loss of all life. If this occurs once per month worldwide, that implies there are only 2500 airliners world wide. According to Wikipedia, United Airlines alone has nearly 400. (UA is just a randomly chosen large airline.)
Or putting the numbers another way, with 400 airplanes, United Airlines should be crashing (with 100% fatality) nearly two per year. Again from Wikipedia, the last such crash was 2001-09-11, and prior to that 1991.
I have heard that some of the early flying boats had horrific safety records, with mean-time-to-crash on the order of 10 to 20 years, but I couldn't find any evidence to back up this memory.
@Ben 7 July 14:30: That's interesting. So, 50% of deaths is third in row, after premature birth and birth defects. So, those two account for more than 50% of deaths EACH, right ..?
Minnesota babies have a 150%+ chance to die. Oops, one slipped through and came up with the newspaper article. Or so.
My thoughts go to the parents that lost their newborn and now get this BS over them.
With regards to stats on aircraft crashing.
They have always been somewhat hyped.
Years ago when doing stats as part of my education the maths instructor showed how it was fairly easy to miss use stats with an example that involved the risk of dying on an aircraft if there was a rock star or other famous person on board the same flight.
The simple fact is that if we have an idea (hypothosis) it can be fairly easy to lead yourself down the wrong path with stats when small sample sizes and several non related or random events happen.
The last time I looked SIDS is something that cannot be tested for medicaly as there is no identified cause. It is basicaly a catch all for babies that just die whilst asleep.
It appears to show great variation depending on where you are in the world, but this is just as likley to be due to reporting norms prevalent in each area.
In the UK we are a little sensitive about childrens deaths due to mothers being prosecuted and convicted or others having their children taken away on nothing more than an "expert witness" with an agenda and the secrecy surounding "family courts".
People need to remember that unless there is hard (factual/testable) evidence you should treat many events as being simply unexplained and not attribute intent or malice to them.
When someone takes booze and pills, gets into a car and kills someone else, do the papers say "Car driving is the no. 1 cause of traffic deaths"?
Taking care of an infant while being drunk is a bad idea anyway, co-sleeping or not.
Jay's comment at July 8, 2009 2:52 AM is neat: the third highest cause of deaths accounts for 50% of them, making the top 3 causes of deaths over 150% of total deaths!
More mundanely, and not vastly accurately because of differing years of the data (2003, 2008 and perhaps others), I contribute the following.
The population of Minnesota is approx 5.2 million (http://en.wikipedia.org/wiki/Minesota ). Annual births per 1,000 population is 13.8, so around 71,700 births per year (http://www.statemaster.com/graph/hea_bir_rat_per_100-birth-rate-per-1-000 ). Infant mortality rate is actually the joint 6th lowest in USA at 5.3 per 1,000 live births, so around 380 infant deaths per year (http://www.statemaster.com/graph/hea_inf_dea_rat-health-infant-death-rate ).
So the 20 or so deaths per year from "suffocation when sleeping with their caretakers" is a bit over 5% of the total for infant deaths, and not the 50% reported as stated my Ms Thompson (unless suffocation or other is material to the statistics).
As one might expect, I've not found statistics on infants sleeping with carers or alone.
Even so, surely a bit of publicity and childcare advice is worth it, to try and reduce that 5% to a much lower figure.
@Michael Woodhams: "How many airliners crash with loss of all life per year? If we're talking 737/A320 or larger, such an event is world news. From my unreliable memory, I'd say that is about 2-4 per year, but for arguments sake, lets call it 12 - one per month."
According to the database airdisaster.com (not counting highjacks):
38 B737 failed and killed more than 90% of their passengers. 23 others failed and killed more than 10%. 22 others failed, destroying the aircraft or some lifes.
3 A319/A320/A321 failed and killed more than 90% of their passengers. Another failed and killed more than 10%. 6 others failed, destroying the aircraft or some lifes.
According to wikipedia, approximatively 6000 B737 were ever produced, and 3800 A319/A320/A321 were ever produced.
So the figure "14%" is not accurate.
"By and large if you are a healthy non smoker then sleeping with a baby is not to much of a problem"
Usually, your posts are well though out and factually correct however, I must take issue with your assumption that third-hand smoke is in some way, going to affect a baby. I would love for you to cite in this instance.
"17% of statistics are made up on the spot"
John "There are more people alive today, than have ever died"
Then is the corollary that from now on more people will die then ever before?
So, going forward, we live forever in increasingly risky times?
I think we're missing the point here. Bruce has given us an exemplar but his thesis
is that people can't calculate risk becuase they don't have enough information
Leaving Mass Media aside (and an entire module of my statistics class was dedicated to media misuse of statistics, for every Bob Woodward there are 1000 Cletus's; (hmmm for our Commonwealth colleagues a Cletus is at about the same level as a pikey, but not that smart. Identifiable by their slack jaws and mouth breathing.)
I learned about risk management 'cause that's what the federal gov't says they manage their IT assets with.
But had to I puzzle it out on my own. No classes, just guesses, a lot of arguing (hey it's billiable), ocasional flashes of insight.
How do we teach how to evaluating risk?
@wsinda "When someone takes booze and pills, gets into a car and kills someone else, do the papers say 'Car driving is the no. 1 cause of traffic deaths'?"
That is absolutely brilliant. I will have to remember that.
Off topic @Piper "Well, actually just Mommy and Junior in the big king bed now. I got banished to a mattress on the floor :-( We do it because the boy won't sleep any other way."
Have you considered taking one side off a crib and attaching it as a "side-car" to the bed? That way you all fit together, with mom next to the baby. That's what we ended up doing, and it worked great for us.
Also, I can't claim to speak for all mothers, but after I had my daughter, I slept so lightly that every little sound or movement woke me. Even mostly asleep, I was aware of where she was in relation to me. I cannot imagine rolling over onto her and not noticing.
We favored putting a large cheap mattress on the floor. So if anybody fell off, they wouldn't have so far to go. Then you can easily add mattresses as needed. (Just don't let the kid fall into (get stick in) the crack.)
You see a lot of this sort of bullshit in modern society. I notice they don't take into account how babies *WANT* constant tactile contact with Mommy or Daddy. They won't sleep well without it. Which means Mommy or Daddy won't sleep well either. It's well established how driving tired is *WORSE* than driving drunk.
Yet nobody factors in the kid's development issues, traffic accidents, poor job performance and layoffs, or what-have-you into the equation. It's all "Oh, somebody got drunk and smothered their kid -- Parents don't sleep with your kids." Pure knee-jerk horror bullshit. (And newspapers wonder why people have stopped reading them?)
I mean, look at our culture: We put kids in cribs. They want to sleep next to us. We belt kids into strollers and forget about them. Other cultures, better than our own, use a long length of cloth [sling] to backpack (or front-pack) their kids. Lots of skin contact. They're up high. They can look all around... The kids are so happy!
A little know fact, at least in our culture: Kids are born ready to use a potty. I'm talking 1-day old newborns here. We stick them in pampers until they're 3 or 4 or 5 years old. (Or sometimes, later.) But you can take a newborn, put them on a baby-potty (plastic seat with a pit), MAKE A HISSING SOUND, and if they have to go they'll go, pretty much on demand. A little positive feedback/reinforcement will further improve [cultivate] their abilities.
Why do you think kids spray (urinate) when you take off their diaper? Tens of thousands of years of evolution, and babies are born knowing a few things. Such as triggers on when to piss & poop.
Or to close their eyes if something is heading towards them. (It's a neat experiment. Babies will stare intently if the object misses them even slightly. They only close their eyes if it's going to hit.)
Or to let go of something if the back of their hand is stroked.
Or to clear their face if something (light) is covering it. Yes, even that skill is there at birth. Cosleeping is not as dangerous as some would have you think.
p.s. Anybody got a pet dog or cat? Ever notice how they want to cosleep with us?
@Anonymous at July 8, 2009 4:32 AM
There is no implication of 3rd/2nd hand smoke doing the damage anymore than 2nd/3rd hand alcohol doing the damage.
Its about how heavy you sleep and if a rather large and uncomfortable lump will wake you up...
Smoking makes people a heavy sleepers by virtue of reducing the effective oxygen carrying capacity of the blood. I forget all the details and can't be bothered re finding the citations...
But as a example, smokers get about half the time diving (with compressed air) as non smokers....
Smoking screws up the body in so many ways. And for what gain? I don't get it...
This view that there are more people alive today than have ever died is distinctly dubious.
Look here for world population from 500 AD to 2000 AD: http://en.wikipedia.org/wiki/...
Look here for typical life expectancies over the ages: http://en.wikipedia.org/wiki/...
Make a very generous assumption of life expectancy of 50 years; ie 1/50'th of the population dies every year prior to 1900. Then integrate by eye, ruler and pencil from 500 to 1900; my figures for the number of deaths are as follows.
From 500 to 750: 250 years times 200 million average population divided by 50 for proportion dying per year gives total deaths in period of 1.0 billion.
And so on for 750..1000 AD: 1.145 billion; 1000..1250 AD: 1.655 billion; 1250..1500 AD: 1.815 billion; 1500..1750 AD: 2.875 billion; 1750..1900 AD (only 150 years): 3.288 billion. Add it all up and get 11.778 billion dying between 500 and 1900 AD. And we have not even included the 20th century deaths.
Compare in excess of 11.8 billion people who have died prior to 1900 AD with the current world population of a bit over 6.5 billion.
In the age of mis-information, there's a real desire for accurate and logical presentation of data by consumers (not politicians or others hoping to win on the basis of people's fear). Is there a higher education degree in the science of presenting relevant data that attempts to be as truthful as possible? Statistics or otherwise?
@Michael Woodhams: ' "Very Reliable Source" is very unreliable.' you should have heard some of his other factoids, especially the ones concerning WMDs in Iraq, American responses to same, and integrity of various American Federal administrations.
@AC2 "17% of statistics are made up on the spot." Heh, heh. Ok, end of my unproductive posts.
@D, I agree completely regarding the contact issue. Constant contact is one of the main ideas behind attachment parenting. My wife's a bit more into AP than I am (I was pretty skeptical at first and still am a bit), but I do see some benefits. I know the plural of anecdote is not data, but when I look at my friends' children and our nephews (no nieces in the family), the ones whose parents are following AP are happier and calmer than the others. Yeah, yeah, correlation != causation and all that, but my observation is that AP seems to work well.
Another AP thing is to wear a wrap with the child in it, rather than dump them in a stroller. Some friends of ours go dancing, and the mom would have the baby in a wrap facing outward from her chest. The parents got to dance, and the baby always seemed to enjoy it, too.
Regarding risks with co-sleeping, I'm surprised no one has mentioned that there are products available that are designed to reduce the risk. E.g., the Snuggle Nest (http://www.snugglenest.com/). We used this when our son was a newborn and loved it. It was great for 2 am feedings, and I don't see how you could roll onto your child in this product (even if you were drunk or whatever). From what I can see, it addresses all of the risks associated with co-sleeping.
@Jay at July 8, 2009 2:52 AM
@Nigel Sedgwick at July 8, 2009 4:26 AM
While I agree that Thompson's 50% figure sounds suspect, her statement is not contradictory to the "third most common cause" statement. Re-read the statements as quoted carefully and you will see why.
Later in the article is the claim that: "Statewide, it's the third most common cause of death in babies less than a year old, behind premature births and birth defects." That doesn't seem to correspond with an average annual mortality of 20.
In any event, common sense would cause me to think that co-sleeping is a bad idea. If I'm 90 kilos, one inadvertent toss in my sleep would be terribly bad for an infant. I'd think a separate bed, even if it's only a meter away, would be wiser.
I'm sure you or someone has already said this, but in modern times, it seems that risk analysis and management should be considered a day to day life skill right alongside with making change---er, rats, everyone uses cash registers...
um, how about cooking? No, that's not it, what with instant microwave dinners...
Ok, a life skill right alongside driving... Err, no, can't use driving and skill in the same sentence.
Managing money? Hm.
Ok, somebody make a friggin iPhone app for managing risk and we'll all be hunky dory.
Risks and perception of progress, a powerful concept. Take Obama recent Russia trip.
Getting rid of a few warheads from both USA/Russia, is a BAD idea, and bad security. No real gains, and who wants materials like these moving around?
Change with some things only makes for more problems! A static inventory, and ease of monitoring existing actions and people passing tests, is critical.
Stupid "victory." A sign that things are all just words.
We should demand better in times like these.
Infant death and its causes is very hard to tease out. Consider the case of SIDS, where almost certainly many cases of accidental smothering are reported as SIDS. It seems to me much more likely that a case of accidental smothering in a crib will be reported as SIDS, while a case of accidental smothering in a co-sleeping situation will be more likely reported as exactly that. And now that a stigma against co-sleeping has become widespread in the medical establishment, we'll see a form of confirmation bias where smothering ("due" to co-sleeping) is the *presumptive* cause of death for infants when co-sleeping is involved.
In any case, my point is that even if you had the raw numbers you desire, comparing deaths with or without co-sleeping, the situation is likely more complicated than that and it *still* won't tell you want to know.
"So the figure "14%" is not accurate."
Not so fast. The way I read that statement is that, for all planes that have taken their last flight, that flight ended in a disaster for 14% of them. Obviously that is true of the Air France and Yemeni aircraft. So they make up part of the 14%. The other 86% end up in places like this:
All those other aircraft which are still flying don't count *because* they are still flying, and haven't yet taken their final flight.
Edit to add: I don't know whether the 14% is true or not (especially with the 100% casualty rider, which I assume is a bit of hyperbole) but looked at that way it certainly isn't completely unreasonable.
PackagedBlue, I warned you before to stick to a single name when you post to this blog. Now you're using the names "Anonymous Coward" and "Anonymous Coward E," and it looks very much as if you're using them specifically when you want to criticize other commenters or take a thread off topic. This is your last warning: do not comment under any name but PackagedBlue unless you want all your comments to be removed.
I'm also going to ask you one more time to communicate more clearly. You're doing much better than when you first showed up here, in that most of your comments are at least partially comprehensible, and some are even on-topic and interesting. But you still lapse far too often into a disjointed, stream-of-consciousness style that nobody should be expected to try to decipher. You've used up all your leeway on this blog, so if you want your comments to stay published, take the time to organize your thoughts before you post.
@Jon: As I stated, my estimate was based on 737/A320 or larger airliners. In the list you link to, there are three such crashes, one of which had only a few deaths (and only one meeting the '100% fatal' criterion). If you count regional airliners (say 25+ passengers) and military flights with significant number of people on board, there are another 5 - but then the number of candidate airliners (denominator in the 14%) becomes very much larger.
For the 2008, I count 7 such crashes, one with 100% fatalities, two with most lives lost, three with no or few lives lost, and one intermediate (32 of 214.)
The original claim "14% of all passenger aircraft ever built have crashed with loss of all lives" would be dominated by 2-4 seater light aircraft, for which data is harder to find. (Although note that "ever built" clearly includes ones which are still flying.)
Ben at July 8, 2009 11:03 AM makes a good point, that I fully accept: the third most common cause relates to 'statewide' infant deaths and the 50% relates to "suffocation when sleeping with their caretakers" in 'the county'.
However, as Ben acknowledges, there remain questions about the reliability of the 50% figure.
Having re-read the newspaper article, I hope carefully enough this time, there is more to be said. I would like to know which county. Is it Dakota Country (where the latest reported death occurred) or is it Hennepin County (that of the place of work of Dr Linda Thompson who gave the 50% figure).
This does actually matter. Hennepin country has a population of over 21% of the whole state, at 1.141 million (2008 estimate), whereas Dakota Country has a population of 0.388 million (2006 estimate). Thus, given statewide annual infant deaths of around 380 and for "suffocation while sleeping with their caretakers" of 20, the figures for Dakota County are much more likely to be affected by random variation within a small sample than for Hennepin County. It is also possible that there are demographic differences between the county and the state that explain all or some of the differences. Finally, there remains the possibility of error in reporting of these statistics.
As Bruce stated in his original posting, and despite my searching out of some of the additional statistics he (and I'm sure many of us) see as lacking from the article, there is a serious problem in newspapers (and/or local officials) splattering around statistics of dubious provenance (and hence of little real use for governmental policymaking).
The best thing of all would be for Dr Linda Thompson to clarify her 50% statistic, and perhaps provide such additional relevant statistics as are known to her. As a paediatrician and member of an official state committee, she surely must have available to her (even if it is not high up in her own personal skillset) advice on what are the meaningful statistics on this issue, including the actual counts as well as percentages.
Michael Shimniok "a friggin iPhone app for managing risk"
Carry around Secure Info's RMS or Trusted Agent Fisma to do risk management on the run? Bleech!
I think a journalist probably "corrected" 5% to 50%, because they thought the real figure looked too low and must be a typo, so please don't assume the doctor was wrong.
@D: You obviously need to stop poking babies in the eye.
The terms of service are understood.
@ Michael Woodhams
Good points, all. Given how poorly the orignial factoid was phrased, I'd be quite happy to argue either side f this particular coin. However, I suspect there is an element of truth in it, regardless of how mangled it's become through multiple retellings.
Yesterday out of interest - and boredom! - I scraped that database for everything this century. The total is 587 crashes, or one per week, each killing somewhere from 0% to 100% of those on board (plus some number on the ground, which I'm ignoring for the purposes of this).* The US seems to feature abnormally highly, but I figure that is probably a result of just a sheer much greater number of flights, coupled with better reporting than say Nigeria or Nepal. The criteria that database uses for inclusion is:
# All civil and commercial aviation accidents of scheduled and non-scheduled passenger airliners worldwide, which resulted in a fatality (including all U.S. Part 121 and Part 135 fatal accidents)
# All cargo, positioning, ferry and test flight fatal accidents.
# All military transport accidents with 10 or more fatalities.
# All commercial and military helicopter accidents with greater than 10 fatalities.
# All civil and military airship accidents involving fatalities.
# Aviation accidents involving the death of famous people.
# Aviation accidents or incidents of noteworthy interest.
Granted the last three probably skew the data a little (although I suspect there haven't been too many airship incidents in the last 10 years), as do the cargo flights, but otherwise it seems pretty comparable to the statement "all passenger aircraft". There doesn't seem - to me - to be any particular reason to arbitrarily cut it off at B-737s, or commuter a/c, or whatever.
So, having said all that, I would not be at all surprised if the correct, or accurate, version of the factoid was something like "10-20% of commerical and/or passenger a/c end their flying careers in a crash, which may kill up to 100% of all on board." That's not as catchy as the original though, heh. Similarly, I wouldn't be too surprised to find out that, say, 20-40% of all cars finish their driving careers with a crash (for some definition of 'car').
* Interestingly, the trend line for the number of fatalities over time is essentially flat. According to excel (rolleyes) the equation of the trendline is y = -5E-05x + 20.995
Bruce, it's not just important to compare the test population (co-sleepers) to a control group (non-co-sleepers). It's necessary to measure the difference, and test whether this difference is statistically significant. It's very unlikely that the death rate is exactly the same in both groups, but the result only matters if it's statistically significant. (Which probably means a chi-squared test in this case.)
The reason for cutting at 737 or bigger was simply ease of data collecting. This is about where 'everybody dies' crashes become world news, so that I could estimate annual rates by inspecting my memory rather than doing actual research.
Not to take this aviation thing too far, 14% is not an absolutely implausible figure (see my first post in this thread). Very hard to prove or disprove!
But the list of 28 crashes in 2009 certainly includes aircraft without passengers, and also includes military ships that don't fit the USUAL meaning of passenger aircraft. Again, almost impossible to prove or disprove.
It seems to me that a considerable majority of paying-passenger (airline, charter, etc.) aircraft crashes have at least one survivor, and for this reason alone, the 14% figure sounds awfully high. I would bet that this claim is hooey ... but I don't know how to prove that!
"At least 71% of Americans alive today will live past age 80" (just made this up, total BS). How does one go about proving or disproving this, without waiting until 71% of us have either died or surpassed age 80? (I'll probably be dead long before, so I'm not taking much of a risk of an "I told you so") It is easy to contrive junk statistics.
@ Michael Woodhams:
Hehe. Fair enough. The problem is, though, that it understates things by an order or magnitude or so :) Also, even within the parameters you set I'll bet there are a *bunch* of large a/c crashes that go more-or-less unreported. For example, the following a/c have crashed in the DR Congo over the last 10 years:
7 x Antonov AN-26
4 x AN-12
2 x AN-28
2 x AN-32
1 x AN-8
1 x Beechcraft 1900
1 x Boeing 737
2 x Cessna 208 Caravan
1 x Ilyushin Il-76
4 x Let-410
1 x DC-9
That's 26 a/c, all of which can hold 16 on more on board, most of which you probably didn't know about (I certainly didn't)
Oh, I think that the factoid *as*stated* is almost certainly wrong. On the other hand, I also think that re-stating it as something like 10-20% of commerical and/or passenger a/c end their flying careers in a crash, which may kill up to 100% of those on board" is quite plausible. Factoids - like Chinese Whispers - have a way of changing in the retelling :)
On topic: here is an analysis of US stats for infant death from co-sleeping in relation to cot sleeping.
"This result shows that it was actually less than half (42 percent) as risky, or more than twice as safe, for an infant to be in an adult bed than in a crib. Based upon these calculations using the CPSC’s own data, we can say that crib sleeping had a relative risk of 2.37 compared with sleeping in an adult bed."
There are issues with the quality of the statistics for this purpose - at best we can say that CPSC statistics do not support the proposition that co-sleeping is a bad thing.
Schneier.com is a personal website. Opinions expressed are not necessarily those of BT.