When somebody is on trial for a crime, it would be nice if we could always calculate mathematically the probability of guilt.
Instead we have to make do with something like "beyond reasonable doubt" etc.
Anyhow I must log off for now,
...later
If this is meant to be a suspicion that the probability argument isn't the way to go, then you would be correct. It isn't.
The correct way is to track
uncertainty in our experiments. All measurements have uncertainties. Those uncertainties can sometimes be propagated mathematically, other times they have to be estimated through experiments. Example, the three different impact cases NIST ran, reflecting uncertainties on the severity of impact.
At each stage, you try to compare the outcomes to other observations to help understand the "best fit" of model versus actual occurrence. This was done at numerous stages -- the debris seen at impact, the windows breaking and spread of fires, the bowing seen in perimeter columns, etc. All of these observations were used to help define the "bounding envelope" of possible cases. Where observations don't exist, we perform what is called sensitivity testing, where we vary our models to see how dependent they are on specific values. Whether or not one bowls a strike is extremely sensitive to the initial vector of a bowling ball. Whether or not one survives a fall from an airplane without a parachute is extremely insensitive to the exact velocity or altitude of the airplane. And so on.
Having estimated errors in the inputs, run bounding cases, and compared the experiment at all phases to other observables, we can say with full confidence that the ultimate outcome -- the collapses -- were expected over a wide range of inputs. It wouldn't have made much difference if the aircraft hit at about 480 MPH (as in WTC 1) or 550 MPH (WTC 2). It wouldn't have mattered much if they hit with 10,000 gallons of fuel on board or 20,000 gallons. The exact
time it took each tower to fall is sensitive to such things, but the ultimate outcome is only slightly sensitive.
That's the way it's done in the real world. Probability of occurence is not the right language. The right term is
uncertainty in the experiment. We can turn that uncertainty into a probability in some cases. Rigorously estimating uncertainty for something so gross and ill-defined as
"if I hit a WTC Tower with an airliner, will it collapse" is tricky, since there are so many variables, but qualitative error estimation as done by NIST provides the correct qualitative answer, which is
"it's quite likely."
Anyone who thinks they have a probabilistic argument against, please present it, because I doubt there are any.
But the cause is what my post was referring to, WRT probabilities.
