A very astute engineer in my last class asked something to the effect "you've showed us this rigorous analysis to calculate jitter and system timing budgets ... but isn't this sort of guess work because we're basing it on a Gaussian curve?"
In a sense it is guess work, but it's the same guess work that puts men on the moon. In the end, that's what statistical analysis is - guess work - but it's all we have when faced with problems that cannot be finitely bounded, like clock jitter. Of course it's much more reliable than guessing. Random clock jitter is characterized by a Gaussian distribution, so using standard deviations to calculate jitter budget is the tool of choice and yields reliable numbers. One could argue that in most cases we've "over engineered" the problem. In many cases, that's what we mean by the term "guarnteed by design". Remember Motorola's 6-sigma design? It was burned into my brain after my time there. They where we're trying to ensure zero defects 99.999% of the time. They were recongnizing that at 6 standard deviations from the mean on the curve, the risk of a defect (in our case a stray rising edge of a clock) was sufficiently low (assumed to be zero in fact).
So I guess I'd concede that it is guess work. Very, very clever guesswork.
To quote John Peel, “I never make stupid mistakes. Only very, very clever ones.” Hopefully the class helps you make cleverer mistakes 
