. . . "But limiting the issue to randomization, one can deductively prove that a random sample will represent the statistical properties of the whole to within a deductively quantizable range of error (for example, one can deductively prove, say, that there is a 99% chance that it is between 88% and 92% probable that I own a car)." .