|
|
Thread Tools | Display Modes |
#1
|
|||
|
|||
Jake Olivier has been at it again, about you know what.
I haven't read Jake's latest research, which is a re-examination or
reuse of data collected in previous studies into you know what efficacy. I don't even pretend to fully comprehend the meaning of his results. Basic statistics I grasp, but there are terms here I've not learned about. There was a guy who found that the results of various case controlled studies overestimate the efficacy of you know what. http://www.fietsberaad.nl/?lang=nl&r...+odds+rat ios Has Zeegers work been discredited? If not, I wonder whether Jake's findings are an overestimation as well? Regardless, as far as I can tell, Jake doesn't consider the risk, only the result after an event. -- JS |
Ads |
#2
|
|||
|
|||
Jake Olivier has been at it again, about you know what.
On 9/29/2016 6:36 PM, James wrote:
I haven't read Jake's latest research, which is a re-examination or reuse of data collected in previous studies into you know what efficacy. I don't even pretend to fully comprehend the meaning of his results. Basic statistics I grasp, but there are terms here I've not learned about. There was a guy who found that the results of various case controlled studies overestimate the efficacy of you know what. http://www.fietsberaad.nl/?lang=nl&r...+odds+rat ios Has Zeegers work been discredited? If not, I wonder whether Jake's findings are an overestimation as well? Yes, certainly. But Jake and his cohorts have an excellent publicity machine. Those who point out the weaknesses in his (and related) methods do not. One fundamental fact is, Olivier's meta-study admitted only data from "case-control" studies based on presentations to hospitals. That's been shown to be the type of study most likely to grossly overestimate helmet benefit. Why? Probably because self-selection is unavoidable. The person who crashed the bike (or that person's parent) is almost always the one deciding to seek medical help. Those who are the most risk-averse will be both the ones wearing the helmet, and the one's more likely to go to the ER "just to be sure," even if the crash was mild. Those not wearing helmets are more likely to have lesser insurance coverage, more likely to go to ER only if they're really hurt - and be more likely to have a high blood alcohol content, which is strongly correlated with bad crashes. I think this is why so many case-control studies (and Olivier's meta-case-control study) predict wonderful things for helmet use; but time trend analyses (TBI counts, or TBI counts per rider) show very low benefit, or even detriment resulting from helmet use. Regardless, as far as I can tell, Jake doesn't consider the risk, only the result after an event. Of course! It goes without saying that riding a bike is terribly, terribly risky. It causes almost all the Traumatic Brain Injury in the world! Oh, wait... -- - Frank Krygowski |
#3
|
|||
|
|||
Jake Olivier has been at it again, about you know what.
On 30/09/16 12:57, Frank Krygowski wrote:
On 9/29/2016 6:36 PM, James wrote: I haven't read Jake's latest research, which is a re-examination or reuse of data collected in previous studies into you know what efficacy. I don't even pretend to fully comprehend the meaning of his results. Basic statistics I grasp, but there are terms here I've not learned about. There was a guy who found that the results of various case controlled studies overestimate the efficacy of you know what. http://www.fietsberaad.nl/?lang=nl&r...+odds+rat ios Has Zeegers work been discredited? If not, I wonder whether Jake's findings are an overestimation as well? Yes, certainly. But Jake and his cohorts have an excellent publicity machine. Those who point out the weaknesses in his (and related) methods do not. One fundamental fact is, Olivier's meta-study admitted only data from "case-control" studies based on presentations to hospitals. That's been shown to be the type of study most likely to grossly overestimate helmet benefit. Why? Probably because self-selection is unavoidable. The person who crashed the bike (or that person's parent) is almost always the one deciding to seek medical help. Those who are the most risk-averse will be both the ones wearing the helmet, and the one's more likely to go to the ER "just to be sure," even if the crash was mild. Those not wearing helmets are more likely to have lesser insurance coverage, more likely to go to ER only if they're really hurt - and be more likely to have a high blood alcohol content, which is strongly correlated with bad crashes. Have you read Zeegers paper, link above? The case controlled studies use an assumption of exposure that turns out, according to Zeegers, not to be valid. He gives a mathematical proof and tests this on 3 studies, confirming his hypothesis. -- JS |
#4
|
|||
|
|||
Jake Olivier has been at it again, about you know what.
On 9/30/2016 5:39 PM, James wrote:
On 30/09/16 12:57, Frank Krygowski wrote: On 9/29/2016 6:36 PM, James wrote: I haven't read Jake's latest research, which is a re-examination or reuse of data collected in previous studies into you know what efficacy. I don't even pretend to fully comprehend the meaning of his results. Basic statistics I grasp, but there are terms here I've not learned about. There was a guy who found that the results of various case controlled studies overestimate the efficacy of you know what. http://www.fietsberaad.nl/?lang=nl&r...+odds+rat ios Has Zeegers work been discredited? If not, I wonder whether Jake's findings are an overestimation as well? Yes, certainly. But Jake and his cohorts have an excellent publicity machine. Those who point out the weaknesses in his (and related) methods do not. One fundamental fact is, Olivier's meta-study admitted only data from "case-control" studies based on presentations to hospitals. That's been shown to be the type of study most likely to grossly overestimate helmet benefit. Why? Probably because self-selection is unavoidable. The person who crashed the bike (or that person's parent) is almost always the one deciding to seek medical help. Those who are the most risk-averse will be both the ones wearing the helmet, and the one's more likely to go to the ER "just to be sure," even if the crash was mild. Those not wearing helmets are more likely to have lesser insurance coverage, more likely to go to ER only if they're really hurt - and be more likely to have a high blood alcohol content, which is strongly correlated with bad crashes. Have you read Zeegers paper, link above? The case controlled studies use an assumption of exposure that turns out, according to Zeegers, not to be valid. He gives a mathematical proof and tests this on 3 studies, confirming his hypothesis. Yes, I've read it. But without the publicity machine, the paper has little impact. Olivier and others simply pretend it was never published. -- - Frank Krygowski |
#5
|
|||
|
|||
Jake Olivier has been at it again, about you know what.
Frank Krygowski wrote in
: Yes, I've read it. But without the publicity machine, the paper has little impact. Olivier and others simply pretend it was never published. A variation on the "file drawer problem"? https://en.wikipedia.org/wiki/Publication_bias -- Andrew Chaplin SIT MIHI GLADIUS SICUT SANCTO MARTINO (If you're going to e-mail me, you'll have to get "yourfinger." out.) |
#6
|
|||
|
|||
Jake Olivier has been at it again, about you know what.
On 10/2/2016 10:07 AM, Andrew Chaplin wrote:
Frank Krygowski wrote in : Yes, I've read it. But without the publicity machine, the paper has little impact. Olivier and others simply pretend it was never published. A variation on the "file drawer problem"? https://en.wikipedia.org/wiki/Publication_bias I just finished the book _Bad Science_ by Ben Goldacre. He deals with that problem fairly extensively, especially regarding studies of medicines. As he expressed it, a pharmaceutical company (or a seller of quackery) has a big motivation to publish research showing their new drug is effective, and to fire up the publicity machine to call attention to the studies. But if research shows no benefit, there's no financial incentive to call attention to that; so the research is likely to go unpublished and sit in someone's file drawer. Even if it is published, no publicity machine kicks into gear, so nobody hears about the paper. Even journalists are (perhaps unwittingly) complicit; saying "This study showed no effect" is a non-story and won't sell papers or advertising. As Goldacre noted, a null result really is good information. If nothing else, it can prevent wasting time on further tests. And regarding quackery, it can save consumers money. All of this applies to the dominant bicycling quackery item: helmets. In this case, calling attention to negative results can do even more; it can dissuade legislators from passing stupid laws. -- - Frank Krygowski |
#7
|
|||
|
|||
Jake Olivier has been at it again, about you know what.
On Thursday, September 29, 2016 at 3:36:38 PM UTC-7, James wrote:
I haven't read Jake's latest research, which is a re-examination or reuse of data collected in previous studies into you know what efficacy. I don't even pretend to fully comprehend the meaning of his results. Basic statistics I grasp, but there are terms here I've not learned about.. There was a guy who found that the results of various case controlled studies overestimate the efficacy of you know what. http://www.fietsberaad.nl/?lang=nl&r...+odds+rat ios Has Zeegers work been discredited? If not, I wonder whether Jake's findings are an overestimation as well? Regardless, as far as I can tell, Jake doesn't consider the risk, only the result after an event. -- JS I'm glad you brought that up. What is being said is that you cannot use the same method of determining "risk" in random data and in control group studies. And what these bicycle helmet studies have been using is the method of analyzing risk appropriate for "control group" studies. In random studies you have little to no idea of the actual risks that are being encountered so simply estimating the number of cyclists wearing helmets and those not and counting the difference in percentages injured is meaningless. |
Thread Tools | |
Display Modes | |
|
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
best way to a lower low on a Jake 2-4? | mtb Dad | Techniques | 5 | May 19th 08 07:27 PM |
JI never: Jake Mccrann arrested in Cuba | War Office | Australia | 0 | November 21st 07 06:09 PM |
Jake, you get wise. You get to church! | PiledHigher | Australia | 3 | June 20th 07 10:19 PM |
Jake Hearts Lance | Ed Sporak | Racing | 6 | July 23rd 06 12:09 AM |
FS: Kona Jake the Snake | Scott | Marketplace | 1 | September 3rd 03 01:35 PM |