The Gamma-Beta connection (the "bank-post office" construction) shows that if X ~ Gamma(a, λ) and Y ~ Gamma(b, λ) are independent, then their sum T = X + Y and their ratio W = X/(X+Y) are themselves independent, with T ~ Gamma(a+b, λ) and W ~ Beta(a, b); this derivation also yields the normalizing constant of the Beta distribution. This independence result, unique to the gamma-beta pairing, licenses computing E[X/(X+Y)] as E[X]/E[X+Y] via uncorrelatedness rather than the (generally invalid) ratio-of-expectations shortcut. The lecture then transitions to order statistics: for i.i.d. continuous random variables, the order statistics are dependent (positively correlated) even though the parent variables are independent, and their marginal CDF and PDF can be derived via a binomial-counting argument and an infinitesimal-interval argument, respectively, with the uniform order statistics giving rise to the Beta distribution as a special case.
The Gamma-Beta connection (the "bank-post office" construction) shows that if X ~ Gamma(a, λ) and Y ~ Gamma(b, λ) are independent, then their sum T = X + Y and their ratio W = X/(X+Y) are themselves …