- Posts: 1436
- Joined: Thu Aug 30, 2012 4:17 pm
It would have to be explained to me what statistically is the violation of two unrelated events occurring in succession.
I would be more worried, not less, if it were never the case. The first injury should have nothing to do with the second just as re-rolling the dice has nothing to do with the previous result. The fact that this seldom does occur--but does in fact occur--is aligned with expectation.
Bulk statistics (like "average") are quite useless for explaining single events. There is a fundamental misunderstanding if the inference is that single or non-representative sampled events need to follow a predetermined likelihood at all times. If so, then there is a deep misunderstanding of what they mean. By definition, it is an impossibility.
The argument is a guy with 30 hits on one side of his card = he must have a rate of 10 hits out of every 36 card chances...or 5 of every 18...or...well do we see how this breaks down? Not to mention how boring--no need for any randomness then. Just follow a script.
Bulk stats of discrete variables that are based on homogenous/gaussian/normal distros are only valid in the long run as cases approach infinity. It approaches nonsense as cases approach 0 (or as the distro no longer looks homogeneous/normal).
And we can't mix selective samples because there is inherent selection bias, no matter how much it sucks when you are the "victim".
I'm open that there are clearly things done with the outcomes for certain stated reasons. But to say there is only bias when it most dramatically affects me really will need lots more than has been offered.
Example of bulk stats gone bad even for a continuous (non-discrete) variable: Temperature in outer space. Very High. Yet you would freeze to death. Why? Because temperature is an intensive variable, and is derived from a simplified scalar interpretation of momentum, energy and collisions. Remove the necessary number of collisions (such as in space where there are few, but mostly high speed particles) to ensure temperature is evenly distributed as roughly 3/2 kT, aka average kinetic energy, and you no longer have what we know as temperature, because the distro is not normal and the cases (collisions) are too low.
OK back to Monty Python