Uncle Albert's Cafe and Ǽ-Theory Bistro

Maybe they read this thread ! :-)

More likely they were caned for such obvious errors. I've been chugging my way though Leonard Mlodinow's book The Drunkard's Walk this year. It's not the easiest of reads - but still way better than many other exposes of statistics. I'm at the hardest bit - the Bayesian logic/minefield.

What I was particularly shocked to read of was OJ Simpson's defence lawyer who made full use of the "Prosecutor's Fallacy" - basically the probability denominator becomes 'of those in custody for this offence' rather than some wider subset of the community - only in reverse. He admitted that he had withheld knowledge that would alter a probability calculation ( in an unfavorable direction for his client of course ) but tellingly admitted ( after the verdict of course ) that it is only sworn witnesses that are law bound to 'tell the truth, the whole truth and nothing but the truth'. He said there is no problem at all if he didn't. That's about as amoral as one can get, but seems to be excused under the rubric of 'blind' justice. I guess whether one accepts that as 'good' depends ...... my guess would be upon which end of a given stick you're personally at.

The terrible topic is homicide of a female by the abusive male partner. The proposal that the relevant statistic is 'most abused females are not murdered by their partner' is not the proposition to consider. The one that mattered for that case was that 'for most murdered abused females it is the abusive partner that did it' : 90+ % !! Thus the blurring over of the choice of denominator. No joke. It worked in court. What a dog & pony show it was.

There were many other faults though : crime scene despoiling does not create OJ's DNA, it can only reduce the chance you will measure it. So finding his DNA despite contamination only speaks for a higher reading from an otherwise 'clean' scene. Also why was fleeing from apprehension not seen as a sign of guilt ? Etc .....

Cheers, Mike.

( edit ) To be exact : a year before her death, say, OJ's wife would have been in the 'abused female' category. She later moved to the 'abused murdered female' category - a subset of 'abused females' obtained by subtracting (a) the living ones and (b) those dead but not murdered. The denominator changed/contracted and thus the ratio increased without the numerator ( consisting of Nicole by herself ) changing. It's not rocket science but one can pretend it is for the purpose of a 'reasonable doubt' construction presented to a jury. It's sad that courtroom operations allow this sort of twaddle to pass in as, ironically, 'evidence' ....

Random Chatty Mode : ON

There's a blizzard of stuff cluttering up low earth orbit. The Space Shuttle has a risk/sensitivity to a flake of paint IIRC - if the relative velocities are bad enough - as disclosed by an examination of one of the front facing windows. More of an issue if you're out and about in only a spacesuit. Norad can only reliably track objects of size baseball and above.

Knowledge, in theory at least, reduces uncertainty and hence refines probability calculations. I had that work in reverse to my advantage recently. I bought a Mitsubishi Lancer high end model ( not an Evo thankfully ) but for that specific model the insurer had no data ( ie. accident/claim statistics ) to give a 'true' risk assessment. So they defaulted to the low risk grouping for that car. I've only made three claims in 30 years that gives me a personally low insurance risk rating ( those claims I've made fortunately devolved to another insurer's payout due to 'fault' ) : mainly my road trauma paranoia ( from my considerable experience treating the results in emergency departments as a young doctor ) means I don't drive 'un-necessarily' or 'for pleasure'. I do only about 15,000 km per year, and I behave in the reverse pattern with respect to the usual 'moral hazard' of insurance ( which is the taking of risks because of the presence of a safety net ).

I've previously mentioned pharmaceutical companies generally only obtusely referring to their numerators and denominators for their statistics : the crucial qualifying comments and disclosures that are within their literature require some considerable digging to unearth. I've stopped seeing 'drug reps' in the last few years as I've tired of being almost uniformly presented with attractive young females in less-than-modest dress to 'inform' me of their company's products. Seriously! It is a new and evolving tactic - that many of my middle aged male colleagues have no issue with BTW. [ I feel like one of the few nuclei in the Marsden-Geiger experiment that sends the alpha particle straight back to source ] Hence the sales pitch can exist at many levels, including the not-quite-so-subliminal. One wonders who they provide to the female practitioners ..... I have no data on that. If push comes to shove/complaint then no doubt I could easily be marked as over-reacting - eg. I'm a 'dirty old man' - by the usual method of plausible deniability. Who would believe me in these days of judgment by label, or trial by media ?? Clever marketing, huh ? :-) :-)

Cheers, Mike.

Buy Nothing Day

wonder if thats covered under the warranty (tongue in cheek) :)

When I get a spare five minutes I'll open it up and photograph it's guts, and post some links here. Let's see what's under the hood! :-)

After all, the data has been removed - so I can autopsy it now. The shattered screen does actually emit some bursts of light at times, but no Apple logo though .....

Cheers, Mike.