Put . Then
Uniqueness of the probability generating function therefore gives
This is Bernoulli thinning of a geometric count, using the convention of counting failures before success.
For an independent nonnegative integer-valued count and identically distributed independent summands, the random sum probability generating function is . Conditioning on proves this, since the conditional probability generating function is .
Here , so
This identity holds throughout , and its power-series extension has radius .
Take independent Bernoulli random variables of success probability , independently of . On the event , the random sum has the binomial distribution with parameters , including the empty sum zero at . It therefore has exactly the specified joint distribution with .
Thus one may realize on a suitable probability space. This is a distributional construction; it does not require a pre-existing to be independent of .
Differentiate the probability generating function inside its radius of convergence:
Since these are and , respectively,
With independent Bernoulli trials of success probability , counts the failures before the first success. There are failures followed by one success with probability , giving the stated geometric distribution on . It counts failures, rather than the total number of trials including the success.
The equilibrium of the first-order linear recurrence is . Subtracting it gives
Consequently
For , interpret the formula at through the initial condition: , while for . For every , .
Apply the law of total expectation to the partition of a sample space given by the possible values of . Since the conditional mean is affine,
Here . This uses linearity of expectation and does not assert independence of successive urn counts.
Only upward and downward moves change the value, so the conditional expectation is
Thus .
Conditional on , urn 1 has red and blue balls, while urn 2 has red and blue balls. The independent selections therefore give
Every other conditional probability is zero. The out-of-range terms at or already have zero probability. These are the transition probabilities of the Bernoulli-Laplace diffusion model.
The first swap necessarily removes one red ball from urn 1 and replaces it with a blue ball. Thus with probability one.
A partition of a sample space is a finite or countable family of pairwise disjoint events whose union is the whole sample space. If is integrable, the law of total expectation states
This follows by summing over the partition of a sample space; absolute integrability justifies the countable sum. For nonnegative , the identity also holds with infinite expectations. Zero-probability parts contribute zero without requiring a conditional value on them.
On the constraint , substitute to obtain
Therefore the two global minima are
Every pair is allowed on either smooth branch , so these are also the only local extrema. There is no local maximum or global maximum: the displayed positive quadratic in has no other stationary point and is unbounded above. At the minima , consistent with the hypothesis in part i.
Assume are continuously differentiable near the point and . The implicit function theorem supplies a local function with
At an interior extremum of , the multivariable chain rule gives
Multiply these identities by and , respectively, and subtract. The terms containing cancel, giving
No assumption is needed for this elimination; dividing by would unnecessarily exclude valid cases. If , this particular local graph argument is unavailable and a different chart or the full Lagrange multiplier condition is needed.
The tangent addition formula gives
Hence . The restriction excludes the poles of tangent that can arise in the possible range . One should not replace this identity by globally: the principal inverse tangent can differ from by an integer multiple of .
The partial derivatives are
Thus the Jacobian determinant is
This vanishing reflects the functional dependence found next.

There are unlisted articles, also show them or only show them.