richlyon wrote:
prof beard wrote:
Rich - you fixation with quantitative evidence is dangerous.
Professor, I have no fixation with quantitive evidence. I have a fixation, if that is what you would call it, with logical consistency applied to a high standard. Have you not? Shouldn't Paul? I sense I am trespassing on both your - and my - patience when I repeat, for the third time, that the first stage of review is to check logic. This is a mechanistic stage that requires no subjective knowledge of the subject matter.
Please look again at
this example of reasoning employed by SafeSpeed. I've [url=http://www.safespeed.org.uk/forum/viewtopic.php?p=59579#59579]invited
Paul[/url] to explain how he feels it is any different from a 'flu vaccinations kills people' argument.
To be absolutely clear: given the three distinctly different conclusions that can be drawn from any graph of two correlated variables e.g.
(1) that increasing road deaths has prompted the increased use of cameras,
(2) deaths and camera convictions have increased without any causal relationship or
(3) that speed camera convictions have caused the increased deaths,
- what his basis is for selecting option (3), and how whatever reasoning that was employed could not just as easily have been used to select option (1) or option (2), because there
simply isn't sufficient data to decide and everything else is just the product of an
a priori decision.
Would you accept it from one of your students? My professor would have hanged me if I'd tried to turn something like that in.
My point is not that no information exists that can prove the detrimental impact of speed cameras on road safety - there may or may not be. My point is that
this is not it, the fact that this is not it can be trivially clarified by simple mechanical argument checking during the first stage of independent verification process, and until that is done
any reasoning derived from it is meaningless.
OK in the interests of open debate I will critique Paul's Fatalities page. This is done with the following provisos:
1 Although I have a degree in maths and studied statistics to first year honours level as an undergraduate, I make no claims to be a statistician, and have not engaged in maths for many years.
2 I am accepting the figures given and tables presented at face value. I have no reason to assume these are "dubious", and have neither the time or energy to test them out.
My first comment to Paul would be that the page should be clearly dated to indicate when it was first written and/or subsequently updated.
My second comment to Paul would be that the page mentions figures not available at the time of writing which must now be available - it would be valuable to update the page to include these to demonstrate whether or not they support the assertions made.
Regarding logic. I am well aware of the concept of the "spurious correalation" - when I was an undergraduate the favourite example used was the rise in teacher's pay correalated to increases in public drunkeness.
Taken in a vacuum, your (Rick's) suggestion that the assertions:
(1) that increasing road deaths has prompted the increased use of cameras,
(2) deaths and camera convictions have increased without any causal relationship or
(3) that speed camera convictions have caused the increased deaths
and indeed I would add:
4) the increased use of cameras has failed to impact on an increase in fatalities
can all be drawn is valid (even allowing for the fact that cameras were introduced prior to the changes in nfatality patterns).
BUT, and here I feel that you (Rick) are being disingenuous to some extent, "spurious correalation" is normally indicated where two sets of figures are ENTIRELY INDEPENDENT. Speed cameras were introduced SPECIFICALLY as a road safety features and were not, ever, intended to have no relationship/impact on road casualties. It is fair, therefore, to postulate on whether a demonstrated correalation may, or may not, be significant. (If on the other hand one insists that cameras were NOT intended to have a relationship to/with casualty figures, that calls into question their function as presented by government)
My criticism of Paul's presentation would be that he has not made clear enough that he is POSTULATING that an indicted correalation is significant and not spurious, and why he is making that assertion.
He has discussed other possible factors, but should have used this discussion to reject other possible explanations of the figures much more overtly.
My main criticism would be that the arguement would have been strengthened (and clearer) if Paul had taken the tack of:
"something is going on here, there are a number of possible explanations - including that speed cameras have no relationship to fatalities, with the serious implications that would imply - I have rejected these ones for these reasons, and would like to postulate that there may well be a causal relationship betweeen cameras and fatalities. Clearly research is required to clarify if this relatioship does actually exist, but I argue that what I have presented is sufficiently worrying as to make such investigations worth funding"
I suspect Paul feels that is what he was trying to say, but with my critical "head" on I would say he has not fully succeeded.
(Apologies for any typos not spotted)
(Edited to put the "dis" in front of ingenuous)