The conditional step of the one-segment martingale #
The segmentwise decomposition reduced the deterministic-cap increment law — the last structural input of the elapsed occupation theory — to the martingale property of the contribution of a single flow segment. Its deterministic half is the truncated one-segment compensation identity. This module supplies the missing conditional step.
The contribution of segment N splits pathwise into a completed part,
carried by the realized reset once the segment has ended, and an active
part, the partial occupation compensator accumulated while the segment is
still running. Conditionally on the history available at time s, on the
event that segment N is active at s with age u = s − T_N, the holding
time has the integrated-hazard law renormalized by the survivor factor
Sur(u), and the channel has the normalized intensity law. The conditional
means of the two parts at a later time t, with v = t − T_N, are therefore
- completed part:
Sur(u)⁻¹ · ∫_u^v haz(r) · (E[reset] − comp)(r) dr; - active part:
Sur(u)⁻¹ · Sur(v) · comp(v).
The truncated one-segment compensation identity evaluates the first integral
as Sur(v)·comp(v) − Sur(u)·comp(u), so the difference of the two conditional
means is exactly −comp(u), which is the value of the segment contribution at
time s. That is the conditional step, and it makes the single-segment
contribution a martingale; by the segmentwise decomposition the
deterministic-cap increment law follows.
The conditional laws themselves — the two integral identities on events of the
filtration, and the centering of a segment not yet started at s — are
collected in DavisSegmentConditionalStepLaw, an interface a concrete model
verifies from its holding-time and channel laws.
Pathwise splitting of the single-segment contribution #
The fully realized and compensated contribution of segment N, carried
once that segment has been completed.
Equations
- S.davisSegmentCompletedContribution hflow ξ initial g N ω = S.davisRealizedResetIncrement hflow ξ initial g N ω - S.davisSampledResetOccupationCompensator hflow g ξ initial N ω
Instances For
The completed part of the contribution of segment N at elapsed time t:
the realized centered reset, retained once the segment has ended.
Equations
- S.davisSegmentCompletedPart hflow ξ initial g N t ω = if ξ.jumpTime (N + 1) ω ≤ t then S.davisSegmentCompletedContribution hflow ξ initial g N ω else 0
Instances For
The active part of the contribution of segment N at elapsed time t:
the partial occupation compensator accumulated while the segment is still
running.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Jump-time form of the single-segment contribution.
The single-segment contribution is the completed part minus the active part.
The two conditional means #
Conditional mean of the completed part of segment N at age v, given
that the segment is still active at age u: the hazard average of the
centered reset increment over [u, v], renormalized by the survivor
factor at u.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Conditional mean of the active part of segment N at age v, given that
the segment is still active at age u: the partial occupation compensator at
v, weighted by the conditional survival probability from u to v.
Equations
- S.davisSegmentConditionalActiveMean hflow g z u v = (S.davisSegmentSurvivor hflow z u)⁻¹ * S.davisSegmentSurvivor hflow z v * S.davisSegmentResetOccupationCompensator hflow g z v
Instances For
The deterministic core of the conditional step. The difference of the
two conditional means is the value the segment contribution already had at
age u.
The conditional step law #
The conditional step of the one-segment martingale: the conditional laws of the holding time and of the channel, expressed as integral identities on the events of the filtration.
On an event of the filtration at time s on which segment N is active, the
completed and active parts at a later time t integrate to the two
conditional means, and a segment not yet started at s carries a centered
contribution at t.
- adapted (t : NNReal) : MeasureTheory.StronglyMeasurable (S.davisElapsedOccupationSegmentContribution hflow ξ hne initial g N t)
The contribution of segment
Nis observable at every elapsed time. Cumulative jump times are observable.
- integrable_completedPart (t : NNReal) : MeasureTheory.Integrable (S.davisSegmentCompletedPart hflow ξ initial g N t) P
The completed part is integrable.
- integrable_activePart (t : NNReal) : MeasureTheory.Integrable (S.davisSegmentActivePart hflow ξ initial g N t) P
The active part is integrable.
- integrableOn_conditionalActiveMean (s t : NNReal) : s ≤ t → MeasureTheory.IntegrableOn (fun (ω : Ω) => S.davisSegmentConditionalActiveMean hflow g (S.davisStateAfterJump hflow ξ initial N ω) (s - ξ.jumpTime N ω) (t - ξ.jumpTime N ω)) {ω : Ω | ξ.jumpTime N ω ≤ s ∧ ¬ξ.jumpTime (N + 1) ω ≤ s} P
The conditional mean of the active part is integrable on the event that segment
Nis active ats. - completed_conditional (s t : NNReal) : s ≤ t → ∀ (A : Set Ω), MeasurableSet A → A ⊆ {ω : Ω | ξ.jumpTime N ω ≤ s ∧ ¬ξ.jumpTime (N + 1) ω ≤ s} → ∫ (ω : Ω) in A, S.davisSegmentCompletedPart hflow ξ initial g N t ω ∂P = ∫ (ω : Ω) in A, S.davisSegmentConditionalCompletedMean hflow g (S.davisStateAfterJump hflow ξ initial N ω) (s - ξ.jumpTime N ω) (t - ξ.jumpTime N ω) ∂P
Conditional law of the completed part on an active event.
- active_conditional (s t : NNReal) : s ≤ t → ∀ (A : Set Ω), MeasurableSet A → A ⊆ {ω : Ω | ξ.jumpTime N ω ≤ s ∧ ¬ξ.jumpTime (N + 1) ω ≤ s} → ∫ (ω : Ω) in A, S.davisSegmentActivePart hflow ξ initial g N t ω ∂P = ∫ (ω : Ω) in A, S.davisSegmentConditionalActiveMean hflow g (S.davisStateAfterJump hflow ξ initial N ω) (s - ξ.jumpTime N ω) (t - ξ.jumpTime N ω) ∂P
Conditional law of the active part on an active event.
- fresh_centered (s t : NNReal) : s ≤ t → ∀ (A : Set Ω), MeasurableSet A → A ⊆ {ω : Ω | ¬ξ.jumpTime N ω ≤ s} → ∫ (ω : Ω) in A, S.davisElapsedOccupationSegmentContribution hflow ξ hne initial g N t ω ∂P = 0
A segment not yet started at
scarries a centered contribution.
Instances For
The contribution of a segment already completed at s no longer changes.
A segment not yet started at s carries no contribution at s.
On an active segment the contribution is minus the partial occupation compensator at the current age.
From the conditional step to the one-segment martingale #
The single-segment contribution is integrable at every elapsed time.
The conditional step. Under the conditional step law the increment of the single-segment contribution is centered on every event of the filtration at the earlier time.
The one-segment martingale. The conditional step law makes the contribution of a single flow segment a martingale.
Segmentwise conditional step laws give the deterministic-cap increment law, the last structural input of the elapsed occupation theory.
Consistency: the interface is satisfiable #
The conditional step law is an interface, so it is worth recording that it is not vacuous. For the zero test function every ingredient vanishes and the law holds for the full-information filtration; this is the degenerate witness that the hypotheses are mutually consistent.
The expected reset increment of the zero test function vanishes.
The partial occupation compensator of the zero test function vanishes.
The single-segment contribution of the zero test function vanishes.
The completed part of the zero test function vanishes.
The active part of the zero test function vanishes.
Both conditional means of the zero test function vanish.
The conditional step law holds for the zero test function.