Alphabeta Math
TheoremStatement: Literature-sourcedProof: AI-adaptedPipeline-generatedprecheck passjudge pass (gpt-6-luna)
How statement and proof provenance work

The first chip identifies the source of the statement or construction; the second identifies the source of its local proof or verification.

  • Literature-sourced: the exact statement appears in a cited source; only wording and notation differ.
  • AI-adapted: a semantically identical restatement of literature-sourced material, modulo indexing, notation, and boundary cases adopted by the library.
  • AI-generated: a genuinely novel statement formulated by AI, with no source for the claim itself.

These labels describe origin, not correctness: citations and verification chips remain separate evidence.

Influence function and sensitivity of the median

Statement

Let F be a Borel probability law on R, let m=m(F) be its lower median, and assume F(m)=1/2. Suppose that on an open interval I containing m there is a continuous density version f satisfying

F((s,t])=∫stf(u) du(s<t, s,t∈I),f(m)>0.

For each fixed x∈R, the right point-contamination influence function of the lower-median functional is

IF⁡(x;m,F)={−12f(m),x<m,0,x=m,+12f(m),x>m.

Consequently, its gross-error sensitivity is γ∗(m,F)=1/(2f(m)).

Facts & Assumptions

Given: A Borel probability law F, its lower median m, the equality F(m)=1/2, and a continuous local density f on an interval I with f(m)>0. Fix a contamination point x∈R.

[F1]

Gross-error contamination path: Fϵ,x=(1−ϵ)F+ϵδx for 0≤ϵ≤1.

[F2]

The lower median of a real probability law: m(G)=inf⁡{t:G((−∞,t])≥1/2} on its stated domain.

[F3]

Statistical functionals and gross-error influence functions: IF⁡(x;m,F) is the right derivative of m(Fϵ,x) at ϵ=0, when finite.

[F4]

Gross-error sensitivity of a functional: γ∗(m,F) is the supremum of the absolute influence values over the directions where they exist.

Proof

technique · direct
1.1givenalgebra

Since I is open, f is continuous at m, and f(m)>0, choose δ>0 and c>0 so that [m−δ,m+δ]⊂I and f(u)≥c throughout this closed interval. For s<t in it, the density identity gives F(t)−F(s)=∫stf(u) du≥c(t−s)>0. Thus F is continuous and strictly increasing there, with F(m)=1/2.

2.1step 1.1F1F2givenalgebra

Suppose first that x<m, and shrink δ so that m−δ>x. For sufficiently small ϵ>0, the number rϵ=(1/2−ϵ)/(1−ϵ) lies strictly between F(m−δ) and 1/2. Continuity and strict increase give a unique qϵ∈(m−δ,m) with F(qϵ)=rϵ. Since rϵ↑1/2 as ϵ↓0, for every 0<η<δ eventually rϵ>F(m−η), so m−η<qϵ<m and qϵ→m. Since qϵ>x, the mixture cdf at qϵ is (1−ϵ)F(qϵ)+ϵ=1/2. For t<x, the contamination term is zero and, by monotonicity and F(m)=1/2, the mixture cdf is (1−ϵ)F(t)≤(1−ϵ)/2<1/2. At t=x, monotonicity gives F(x)≤F(m−δ)<rϵ, so (1−ϵ)F(x)+ϵ<(1−ϵ)rϵ+ϵ=1/2. For x<t<qϵ, the contamination term is ϵ and F(t)<F(qϵ)=rϵ: below the local interval this follows from monotonicity and F(t)≤F(m−δ)<F(qϵ), and inside it from strict increase. Therefore the mixture cdf at such t is (1−ϵ)F(t)+ϵ<(1−ϵ)rϵ+ϵ=1/2. Together with the cases t<x and t=x, this puts the mixture cdf below 1/2 for every t<qϵ, while it equals 1/2 at qϵ, so [F2] gives m(Fϵ,x)=qϵ.

2.2step 1.1F1F2givenalgebra

If x>m, shrink δ so that m+δ<x. For small ϵ>0, sϵ=1/(2(1−ϵ)) lies strictly between 1/2 and F(m+δ). There is a unique qϵ∈(m,m+δ) with F(qϵ)=sϵ. Since sϵ↓1/2 as ϵ↓0, for every 0<η<δ eventually sϵ<F(m+η), so m<qϵ<m+η and qϵ→m. Now qϵ<x, so the mixture cdf at qϵ is (1−ϵ)F(qϵ)=1/2. For every t<qϵ, the contamination term is zero; if m≤t<qϵ, strict increase on the local interval gives F(t)<F(qϵ), and if t<m then monotonicity gives F(t)≤F(m)<F(qϵ). Thus the mixture cdf is below 1/2 for every t<qϵ, while it equals 1/2 at qϵ, so [F2] gives m(Fϵ,x)=qϵ.

3.1step 2.1givenalgebra

In the case x<m, step 2.1 gives F(qϵ)−F(m)=−ϵ/[2(1−ϵ)] and qϵ→m. The density identity and continuity of f imply [F(qϵ)−F(m)]/(qϵ−m)→f(m). Dividing these two relations yields (qϵ−m)/ϵ→−1/(2f(m)).

3.2step 2.2givenalgebra

In the case x>m, step 2.2 gives F(qϵ)−F(m)=ϵ/[2(1−ϵ)] and qϵ→m. The same density quotient tends to f(m), hence (qϵ−m)/ϵ→+1/(2f(m)).

4.1step 1.1F1F2F3step 3.1step 3.2givenalgebra

If x=m, continuity of F at m gives Fϵ,m(m−)=(1−ϵ)/2<1/2 and Fϵ,m(m)=(1+ϵ)/2>1/2. Monotonicity then puts every t<m below the median threshold, while m itself is above it, so the lower median remains exactly m and its right derivative is 0. By [F3] and steps 3.1–3.2, these three cases give the displayed pointwise influence function.

5.1givenF4step 1.1step 3.1step 3.2step 4.1algebra∎

The absolute influence value is 1/(2f(m)) for every x≠m and zero at m. Both sides of m are nonempty in R, so the supremum over all directions is 1/(2f(m)) by [F4]. Since F has a density near m, F{m}=0; therefore the conventional expression (1/2−1{x≤m})/f(m) agrees with this pointwise derivative F-almost everywhere, though its value at x=m is not the literal right derivative.

Remarks

The conventional indicator formula in the Statement of the controlling design is an F-almost-everywhere representative. At the single point x=m, the right point-contamination derivative of the lower median is 0; retaining this pointwise value makes the definition agree exactly with the computed limit.

Depends on

Used by

Dependency tree · two levels

11 results within two dependency steps of this one, each drawn at its shortest distance from it. An arrow runs from a result to what uses it, so the chart reads left to right and ends at this result, which carries a heavier outline. Every node is a link to that result. Click elsewhere on the chart to enlarge it.

Sources