brain sizes: Einstein's and women's

John Knight jwknight at polbox.com
Sat Oct 19 21:00:08 EST 2002



The Source of the DWI [driving while intoxicated] laws

Not one single American scientists, nor mathematician, nor even student, teacher, politician, nor bureaucrat, has even taken the time to critique the methodology used to justify the "DWI" [read: driving while intoxicated] laws.  If even one competent mathematicain had, government and "we the people" would have noted the serious methodological flaws on which these fraudulent laws are based and done something to PREVENT that draconian laws that followed.

The flaws are so glaring and so obvious and so egregious that it's really an embarassment to each and every honest, concerned American citizen.

There are three studies which are referenced when governments attempt to justify these draconian DWI laws: 

  a.. The Borkenstein study. 
  b.. The "H.P. Krüger" study. 
  c.. The Perrine study. 
All three studies contain serious methodological flaws similar if not identical to the following: 

  a.. SAMPLING DEMOGRAPHICS: "Field operations were conducted on Friday and Saturday nights during two two-hour periods at separate sites, at one site between 10 PM and midnight, and at the other between 1 AM and 3 AM. Data from the 96NRS is representative only of locations and periods when drinking and driving is most prevalent (i.e., not all times or roadways in the 48 contiguous states)."

  b.. ACCIDENT RISK BIAS:  The methodology used to assess accident risk under-represents the expected accident rates of each BAC class.

  c.. DISQUALIFIED CONTROL GROUP:   Key drivers from the control group were excluded from the results, even though including them may have almost doubled the number of drivers with a BAC > 0.

  d.. CONTROL GROUP BIAS:   No reason was given for removing another 85 drivers from the control group, who may have been another 14% of the drivers with a BAC > 0.

  e.. REPORT BIAS:  To determine who was the cause of the accident, they all relied on police who have been trained that drinking alcohol automatically presumes guilt.

  f.. LOGICAL ERROR:   All three almost ignored the most important fact about their study group:  the SAFEST drivers were drivers with a BAC > 0 and < 0.8, which even their own data showed to be 89.8% of all drivers with a BAC > 0.

  g.. DROWSY DRIVING: None of these studies took into account "drowsy driving", which the NHTSA estimates could account for 5% of night time accidents.  This alone could cause even more accidents than they attributed to drivers with a BAC between .18 to .20 and .20 and above, combined.

  h.. ADVOCATE GROUP:   With such obvious advocacy work, the composition of this control group could easily have been manipulated in ways that weren't even reported.

SAMPLING DEMOGRAPHICS

It's impossible to know what effect drinking and driving has on accidents if a complete data set for ALL times of the day and night, all driving conditions, and all other pertinent factors, is not known.   By excluding drivers who may have been "intoxicated" at 7 am in the morning (by their definition of the word "intoxicated"),  it's impossible to detect accidents that may have been alcohol related at other times of the day.  By excluding these other potential "alcohol related accidents", it's mathematically impossible to get a complete and unbiased result for the sample they studied.  This failure alone disqualifies any opinion that might be issued by the scientists who conducted this study.

ACCIDENT RISK BIAS

Kruger's study estimated that the "accident risk" of drivers with a BAC > .02 and < .04 should have resulted in 30 fatal accidents, which is 9 accidents less than their actual number of 21 accidents, indicating that the odds ratio for such drivers was 0.7 (or that they were 30% less likely than the average to cause an accident).  But drivers with this BAC level were 1.71% of the control group, so they should have been 1.71% of the 1,968 drivers in accidents (or 34) included in this study.  The fact that only 21 of the drivers in this BAC class were involved in the accidentsindicates that their odds ratio was actually .62, meaning that they were 38% less likely than average to cause an accident.

This bias caused the following errors:

      bac class
     Kruger
     Proportionate
     Percent Difference
     
      < 0.02
     .96
     0.86
     12.4
     
      < 0.04
     0.7
     0.62
     12.4
     
      < 0.06
     1.08
     0.93
     16.1
     
      < 0.08
     3.38
     3.1
     8.8
     
      < 0.10
     11.5
     8.8
     30.6
     
      < 0.12
     6.0
     5.2
     16.1
     
      < 0.14
     11.5
     11.7
     -2.1
     
      < 0.16
     31
     35.6
     -13.1
     
      < 0.18
     29
     22.2
     30.6
     
      < 0.20
     28
     25.7
     8.8
     
      >= 0.20
     32
     29.4
     8.8
     

In other words, where we would expect the odds ratio for drivers in the .04 to .06 class to be .93 (where they would be 7% less likely than average to cause an accident), Kruger predicted that their odds ratio would be 1.08 (where they would be 8% MORE likely to cause an accident).

DISQUALIFIED CONTROL GROUP

Kruger reports that "Of those asked for a breath sample, 9128 (94.8%) agreed".  BUT ONLY 6.3% OF THE DRIVERS HAD A BAC > 0!  It's far more significant that 5.2% of nighttime drivers would refuse to take a breath sample requested by the police than it is that 6.3% had a BAC > 0.  But this point was entirely ignored.  The opportunity was here to determine what percentage of those who refused to take the test were actually intoxicated, but this opportunity was missed, thus it's now impossible to know with certainty what percentage of them had a BAC > 0.  But it's entirely possible that all of them did, which would have almost doubled the percentage of drivers in the control group with a BAC > 0, from 6.3% to 11.5%.  In addition, due to CONTROL GROUP BIAS, another 85 drivers were omitted from the control group with no explanation, and it's possible that they too had a BAC > 0, which would have increased the percent of drivers in the control group with a BAC > 0 by 14%.   This would mean that the percent of drivers in the control group with a BAC > 0 was actually 12.4%.

The use of such a disqualified control group can never be justified.  The only way to salvage anything from this "study" is to make some wild assumptions about the BAC levels of those who refused the breath test.  What they SHOULD have done is a min and a max using the known data, and making several assumptions about the distribution of the BAC levels.  For example, if it's assumed that the distribution of the unknown 5.2% and the unknown 0.9% was similar to the known distribution of the 6.3%, then drivers with a BAC > 0 and < 0.02 would be 50% less likely to have an accident than a non-drinking driver, rather than only 4% less likely.  Similarly, drivers with BAC > 0.02 but < .04 would be 64% rather than 30% less likely to have an accident, and drivers with BAC > .04 but < .06 would be 46% less likely rather than 5% more likely to.

It doesn't make any sense, though, that drinking one drink and getting a BAC > 002 but < 0.04 would decrease your likelihood of having an accident by 64%, though.  This would put such drivers into Mario Andretti's class, and it's not likely that drinking alcohol is THAT beneficial.   Thus a more reasonable assumption would be that those who refused the test tended to be the drivers with the highest BAC levels, including those with a BAC > .20.   It should be noted that the use of this biased control group resulted in the prediction that only 2 drivers with a BAC > .20 should have been involved in fatal accidents, compared to the actual number of accidents of 64.  This is what caused the scientists to state that such drivers had 62 more accidents than they should have, making them 31 times more dangerous than non-drinking drivers.  But if only 62 or 12.4% of these 501 drivers who refused the test had a BAC > .20, then they would NOT have been over-represented in fatal accidents.

This would drive the conscientious scientist to make some astute assumptions about the potential bias in the way the data was collected, managed, and presented.  It would make him suspicious of this obvious advocacy work, which should lead him to presume that the media hype about drinking and driving may be what encouraged these German scientists to use such a faulty control group in the first place.  It should make him consider the possibility that, of the 9,629 drivers who were reported to be in the control group, 8,438 or 87.6% of them had a BAC = 0 just as reported, whereas 1,191 or 12.4% of them had a BAC > 0 but almost   half of them were omitted for political reasons.

LOGICAL ERROR

By failing to note how much safer drivers with a BAC between 0 and 0.04 are, and that they are almost 90% of all drivers, they failed to estimate how many lives would be saved if ALL drivers had the safety record of these moderately drinking drivers.  

Does anyone on this forum care to venture a guess as to how many lives this "DWI campaign" has cost the country, so far?



John Knight



ps--this can be seen in the original form at http://christianparty.net/dwi.htm
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://iubio.bio.indiana.edu/bionet/mm/neur-sci/attachments/20021020/9f5e0378/attachment.html


More information about the Neur-sci mailing list