(reposted with sage advice)
It would help greatly, if reporters and policy makers understood subject material before they commented on it...(wishful thinking)
Today I found, a report that the odds of dying from an opioid overdose now surpass those of dying in car crash. well we all intrinsically knew that (or should have) but the next lines shocked me....
FWIW, the numbers are 1/96 for opioids vs 1/103 for 'car' crashes
Ok, sounds scary and we should all jump, but the comparo is MEANINGLESS.
Rule:
The odds of being negatively affected by any activity only apply IF you engage in that activity. (this leaves off the non-zero probability[1] of someone holding you down and force feeding you opioids until you die.)
For example, I dont take opioids. Therefore if the death rate was 100%, it affects me not. I do however, drive, so the death rate for being in a vehicle would be justifiably concerning.
You cannot take the # of incidences for a subset of a population, and then calculate the probability using the ENTIRE population and equate them with another subset.
IT also means that the opioid death rate is way scarier than reported. I have seen it stated that 10% of the population uses illegally consumed drugs (meaning drugs w/o prescription and outside guidelines)
So assuming all the drugs taken have an equal fatality rate and opioids are the only drug, the actual death rate is 10x that stated. (or rather 1/9.6)
From the article, they state that over a persons lifetime (pay attention to that word) the chance across the US population of people over 12 is 1 in 96 opioid users will die from them. Well ok, but only 10% of the population (sometimes I am sure it is more) take drugs 'improperly'[2] so really its 1 in 9.6 users will die and THAT has to assume that all users are opioid users - which aint true. (in other words, if you have an opioid problem - get help asap because the true rate is likely higher than that)
I submit that the death rate per 'x' participants has ALWAYS been higher than 'car things'
the article also states that they came up with the numbers (btw - the 'they' is the National Safety Council and they should know better. I fear they dont, but they should...) by taking a yearly chance and multiplying by a persons life expectancy. W. T. H. ?!?!?!?
I dont expect everyone reading to immediately grasp the sheer stupidity of this but lets try to lay out why this is faulty:
In any one year, you kill x many people on the roads in the US. the current number is appx 33000 people per year, based on last year in the US and this number is not a constant, but for simplicity, lets use last year as a constant.
If we assume that ANY us citizen can get into a vehicle in a year, then in a gross sense, using a population of 310M, we get 1 out of every 10000 people will die in a traffic wreck. And each car trip does not influence the rest. So this takes the population and averages it as 'the person who drives exactly 1 time per year has the same chance in a year as the person who drives 730 times a year' (a person who drives to work, and back every day with no break) and while it is true that the chance of ANY trip is the same as ANY other trip (ignoring, weather, traffic, TOD etc) the person who drives a lot simply enters the lottery more times.
Important point 1:
Does entering a DIFFERENT lottery many times influence your outcome? NO.
Important point 2:
Does entering the SAME lottery many times influence your outcome? YES.
Hence, a car crash in the US might claim 1 person, but a megabus crash might claim 10 (and a bus crash in india might claim 250....) so CLEARLY using the number of deaths on a road divided by the number of people who are alive, is a crappy way of finding the truth, and easily manipulable ( which is why auto insurance companies use it)
It was only relatively recently for example, that insurance companies -who use stuff like this to set rates- were force to divorce 'number of miles driven' from 'driven to work'. It used to be that if you drove a lot of yearly miles, you got put into the same hopper as commuters. Many cars in one space increase the odds of a crash
(and I maintain that you have to BE in a crash in order to die from one)
so commuting is riskier than driving. I, work from my home office and do not participate in the twice daily bumper car rally. But at the same time, I average over 26K miles a year on my DD, not including the OTHER cars in the stable (and this number is down from >50K) My insurance company HATED HATED HATED that fact and audited me twice a year to try and catch me driving to work. But PA passed a law and now almost all of my cars are rated at 'pleasure use', which does not mean I am getting pleasure (I wish) but rather the act of driving NOT for work is considered in the industry as 'pleasure' in the sense of 'drive as I please' Important distinction - but I digress.
the next fail: with the NSCs number is they list it as odds across your lifetime, BUT!!!!! the lifetimes are not equal! IF a 2 year old dies in a car crash then that lifetime is OVER. if a 16 year old dies in a crash then that lifetime is OVER. Ditto 40, 50 etc up to the average average of 78. IF their numbers are to have meaning then it implies that a 2 year old can climb out of the car seat and jump up and down while holding metal forks (because 2 << 78) but an 80 year old has got to be shitting themselves driving themselves to market cuz its 'coming' any day now...(by their law, the 103 year old person al Roker visits will die later that day in a car crash all things considered)
and lastly, the most important fail and how I know no true statistician is at the NSC : the population.
yeah there are 310M people NOW and lets assume that birth rate = death rate always such that there were 310M people THEN, 78 years ago. The problem is, they were a DIFFERENT 310M people. If someone reading this is 78 years old - exactly average average (male = 77, women=79, #men=#women) then in each year, for a statistic of 'across my lifetime' to make sense, they have to know the names of everyone alive in each year and never double count anyone.
The population to divide by is now higher. And the odds of dying across your lifetime, just went lower. Yeah its a lot of work to derive a semi useful statistic, given that as said before, one car trip does not affect the outcome of the next one
(think of the coin flipping test)
to put this in better terms, if our chances of dying in any car crash are 1/10000 people and we do NOT take into account lottery entries (each time you start it up) then you have to theoretically drive on average 10000 times that year to 'win', aka die - which equates to just over 27 trips a day, every day. (does this mean UPS delivery men are dying left and right?)
Granted, there is so much not taken into account, those drunk have a better chance of winning, those driving fast, those driving in snow, at night, in deer inefested woods...this part actually makes sense to set car insurance rates...faster cars, more risk, snow belt states, more risk, deer states, more risk, dui convicts...more risk (recidivism among DUI is very high, much higher than known becuz so many get away with it)
so contrast to REAL numbers: in 2017 (the last year so far collected) there were 70000ish drug deaths of which 2/3 are attributed to opioids. roughly 46000. WE know that 32799 traffic deaths came about (33K)
so from the RAW numbers there were appx 35% MORE opioid deaths than traffic deaths. BUT! one source lists the age adjusted deaths from 15-up, another 12 up..
you can die in a car crash on your way home from the birthing hospital, or one the way to the hospital as a geezer - in other words - any age. Plus, it was previously clamed that only 10% of the people 12-whatever use drugs so the other 90% dont get to participate.
If the article was meant to scare us, it does a piss poor job. A lot of things are more dangerous than driving. ATV riding...sky diving...few of us know a traffic fatality personally. And it does a disservice to the drug problem overall as it understates the true danger.
Notes:
[1]= It can be shown that the probability of ANY event, and I mean ANY, is never zero. Meaning, there are odds that a male reading this can not only grow boobs and wings, but also fly over moscow and drop nukes from their toes. It would be the must see TV event of the year....but unlikely
[2] How they classify opioid users is tricky and not well sorted out. You can have people illegally obtaining (without a script) opioids that themselves are illegal (fentanyl - not sanctioned by the DEA)
Or illegally obtaining (sans script) legally produced percocets
Or legally obtaining (script) legal drugs ( percocets) and glomming them all at once.
The whole phenomenon has to be long winded summed up as 'people taking medication outside of the medically intended purpose' or in other words 'not following directions on the label' cuz I doubt CVS or walgreens will ever put on the label "down entire bottle at once"
ps: if it turns out that 100 AK users read this post, STATISTICALLY, 10 of them abuse opioids and of that, 1 will die. 10% of AK abusing opioids? well, that does explain threads about interconnects and power cords...
It would help greatly, if reporters and policy makers understood subject material before they commented on it...(wishful thinking)
Today I found, a report that the odds of dying from an opioid overdose now surpass those of dying in car crash. well we all intrinsically knew that (or should have) but the next lines shocked me....
FWIW, the numbers are 1/96 for opioids vs 1/103 for 'car' crashes
Ok, sounds scary and we should all jump, but the comparo is MEANINGLESS.
Rule:
The odds of being negatively affected by any activity only apply IF you engage in that activity. (this leaves off the non-zero probability[1] of someone holding you down and force feeding you opioids until you die.)
For example, I dont take opioids. Therefore if the death rate was 100%, it affects me not. I do however, drive, so the death rate for being in a vehicle would be justifiably concerning.
You cannot take the # of incidences for a subset of a population, and then calculate the probability using the ENTIRE population and equate them with another subset.
IT also means that the opioid death rate is way scarier than reported. I have seen it stated that 10% of the population uses illegally consumed drugs (meaning drugs w/o prescription and outside guidelines)
So assuming all the drugs taken have an equal fatality rate and opioids are the only drug, the actual death rate is 10x that stated. (or rather 1/9.6)
From the article, they state that over a persons lifetime (pay attention to that word) the chance across the US population of people over 12 is 1 in 96 opioid users will die from them. Well ok, but only 10% of the population (sometimes I am sure it is more) take drugs 'improperly'[2] so really its 1 in 9.6 users will die and THAT has to assume that all users are opioid users - which aint true. (in other words, if you have an opioid problem - get help asap because the true rate is likely higher than that)
I submit that the death rate per 'x' participants has ALWAYS been higher than 'car things'
the article also states that they came up with the numbers (btw - the 'they' is the National Safety Council and they should know better. I fear they dont, but they should...) by taking a yearly chance and multiplying by a persons life expectancy. W. T. H. ?!?!?!?
I dont expect everyone reading to immediately grasp the sheer stupidity of this but lets try to lay out why this is faulty:
In any one year, you kill x many people on the roads in the US. the current number is appx 33000 people per year, based on last year in the US and this number is not a constant, but for simplicity, lets use last year as a constant.
If we assume that ANY us citizen can get into a vehicle in a year, then in a gross sense, using a population of 310M, we get 1 out of every 10000 people will die in a traffic wreck. And each car trip does not influence the rest. So this takes the population and averages it as 'the person who drives exactly 1 time per year has the same chance in a year as the person who drives 730 times a year' (a person who drives to work, and back every day with no break) and while it is true that the chance of ANY trip is the same as ANY other trip (ignoring, weather, traffic, TOD etc) the person who drives a lot simply enters the lottery more times.
Important point 1:
Does entering a DIFFERENT lottery many times influence your outcome? NO.
Important point 2:
Does entering the SAME lottery many times influence your outcome? YES.
Hence, a car crash in the US might claim 1 person, but a megabus crash might claim 10 (and a bus crash in india might claim 250....) so CLEARLY using the number of deaths on a road divided by the number of people who are alive, is a crappy way of finding the truth, and easily manipulable ( which is why auto insurance companies use it)
It was only relatively recently for example, that insurance companies -who use stuff like this to set rates- were force to divorce 'number of miles driven' from 'driven to work'. It used to be that if you drove a lot of yearly miles, you got put into the same hopper as commuters. Many cars in one space increase the odds of a crash
(and I maintain that you have to BE in a crash in order to die from one)
so commuting is riskier than driving. I, work from my home office and do not participate in the twice daily bumper car rally. But at the same time, I average over 26K miles a year on my DD, not including the OTHER cars in the stable (and this number is down from >50K) My insurance company HATED HATED HATED that fact and audited me twice a year to try and catch me driving to work. But PA passed a law and now almost all of my cars are rated at 'pleasure use', which does not mean I am getting pleasure (I wish) but rather the act of driving NOT for work is considered in the industry as 'pleasure' in the sense of 'drive as I please' Important distinction - but I digress.
the next fail: with the NSCs number is they list it as odds across your lifetime, BUT!!!!! the lifetimes are not equal! IF a 2 year old dies in a car crash then that lifetime is OVER. if a 16 year old dies in a crash then that lifetime is OVER. Ditto 40, 50 etc up to the average average of 78. IF their numbers are to have meaning then it implies that a 2 year old can climb out of the car seat and jump up and down while holding metal forks (because 2 << 78) but an 80 year old has got to be shitting themselves driving themselves to market cuz its 'coming' any day now...(by their law, the 103 year old person al Roker visits will die later that day in a car crash all things considered)
and lastly, the most important fail and how I know no true statistician is at the NSC : the population.
yeah there are 310M people NOW and lets assume that birth rate = death rate always such that there were 310M people THEN, 78 years ago. The problem is, they were a DIFFERENT 310M people. If someone reading this is 78 years old - exactly average average (male = 77, women=79, #men=#women) then in each year, for a statistic of 'across my lifetime' to make sense, they have to know the names of everyone alive in each year and never double count anyone.
The population to divide by is now higher. And the odds of dying across your lifetime, just went lower. Yeah its a lot of work to derive a semi useful statistic, given that as said before, one car trip does not affect the outcome of the next one
(think of the coin flipping test)
to put this in better terms, if our chances of dying in any car crash are 1/10000 people and we do NOT take into account lottery entries (each time you start it up) then you have to theoretically drive on average 10000 times that year to 'win', aka die - which equates to just over 27 trips a day, every day. (does this mean UPS delivery men are dying left and right?)
Granted, there is so much not taken into account, those drunk have a better chance of winning, those driving fast, those driving in snow, at night, in deer inefested woods...this part actually makes sense to set car insurance rates...faster cars, more risk, snow belt states, more risk, deer states, more risk, dui convicts...more risk (recidivism among DUI is very high, much higher than known becuz so many get away with it)
so contrast to REAL numbers: in 2017 (the last year so far collected) there were 70000ish drug deaths of which 2/3 are attributed to opioids. roughly 46000. WE know that 32799 traffic deaths came about (33K)
so from the RAW numbers there were appx 35% MORE opioid deaths than traffic deaths. BUT! one source lists the age adjusted deaths from 15-up, another 12 up..
you can die in a car crash on your way home from the birthing hospital, or one the way to the hospital as a geezer - in other words - any age. Plus, it was previously clamed that only 10% of the people 12-whatever use drugs so the other 90% dont get to participate.
If the article was meant to scare us, it does a piss poor job. A lot of things are more dangerous than driving. ATV riding...sky diving...few of us know a traffic fatality personally. And it does a disservice to the drug problem overall as it understates the true danger.
Notes:
[1]= It can be shown that the probability of ANY event, and I mean ANY, is never zero. Meaning, there are odds that a male reading this can not only grow boobs and wings, but also fly over moscow and drop nukes from their toes. It would be the must see TV event of the year....but unlikely
[2] How they classify opioid users is tricky and not well sorted out. You can have people illegally obtaining (without a script) opioids that themselves are illegal (fentanyl - not sanctioned by the DEA)
Or illegally obtaining (sans script) legally produced percocets
Or legally obtaining (script) legal drugs ( percocets) and glomming them all at once.
The whole phenomenon has to be long winded summed up as 'people taking medication outside of the medically intended purpose' or in other words 'not following directions on the label' cuz I doubt CVS or walgreens will ever put on the label "down entire bottle at once"
ps: if it turns out that 100 AK users read this post, STATISTICALLY, 10 of them abuse opioids and of that, 1 will die. 10% of AK abusing opioids? well, that does explain threads about interconnects and power cords...