Conditioning on a set of non-zero measure yields a probability measure
ProvedProbabilityTheory.cond_isProbabilityMeasure_of_realConditioning a measure on a set of non-zero real measure gives a probability measure.
For a measure on and a set with , the conditional measure
is a probability measure, i.e. it assigns total mass .
The hypothesis is stated with the real-valued measure rather
than , and this is the useful form: Measure.real sends to , so
simultaneously rules out and .
Both exclusions are necessary — conditioning on a null set is undefined, and conditioning on a
set of infinite measure does not normalise — so the single real-valued hypothesis captures
exactly the right condition where the ENNReal version would need two.
This is the standing hypothesis-discharging lemma in probabilistic arguments: once
is known to be a probability measure, the whole IsProbabilityMeasure
API (total mass, expectations, entropy) becomes available for it.
Formalization note. μ[|s] is Mathlib's notation for ProbabilityTheory.cond μ s, and
μ.real s is (μ s).toReal.
import Mathlib
namespace ProbabilityTheory
open MeasureTheory ProbabilityTheory in
theorem cond_isProbabilityMeasure_of_real {α : Type*} {_ : MeasurableSpace α} {μ : Measure α}
{s : Set α} (hcs : μ.real s ≠ 0) :
IsProbabilityMeasure μ[|s] := by sorry
end ProbabilityTheory