Plurality, bias, α and γ of a color distribution (Section 3) #
For a k-color distribution c : Fin k → ℕ with ∑ₕ cₕ = n the paper uses
m(c) = maxₕ cₕ(maxc) andM(c) = {j | cⱼ = m(c)}(argmaxSet);- the bias
s(c) = m(c) - max_{h ∉ M(c)} cₕif the plurality color is unique and0otherwise (bias); α(c) = (n - m(c)) s(c) / n²andγ(c) = (n m(c) - ∑ₕ cₕ²)/n² - α(c);μⱼ(c) = cⱼ (1 + (n cⱼ - ∑ₕ cₕ²)/n²), the expected next count of colorj(Lemma 2.1,Plurality.expected_count).
Lemma 3.1 bounds these quantities and Lemma 3.2 rewrites μⱼ in terms of
them. Everything here is elementary algebra about finite sums of naturals.
m(c): the plurality size.
Equations
Instances For
M(c): the plurality colors.
Equations
- Plurality.argmaxSet c = {h : Fin k | c h = Plurality.maxc c}
Instances For
The largest count among the colors outside M(c) (0 if there is none).
Equations
- Plurality.secondc c = {h : Fin k | c h ≠ Plurality.maxc c}.sup c
Instances For
s(c): the bias, i.e. the gap between the plurality color and every other
color when the plurality color is unique, and 0 otherwise.
Equations
- Plurality.bias c = if (Plurality.argmaxSet c).card = 1 then Plurality.maxc c - Plurality.secondc c else 0
Instances For
α(c) = (n - m(c)) s(c) / n².
Equations
- Plurality.alpha n c = (↑n - ↑(Plurality.maxc c)) * ↑(Plurality.bias c) / ↑n ^ 2
Instances For
γ(c) = (n m(c) - ∑ₕ cₕ²) / n² - α(c).
Equations
- Plurality.gamma n c = (↑n * ↑(Plurality.maxc c) - ∑ h : Fin k, ↑(c h) ^ 2) / ↑n ^ 2 - Plurality.alpha n c