The Itô integral of simple nonanticipating integrands (Gardiner §4.2.1, §4.2.4, §4.2.6 e–f) #
A simple integrand (Gardiner.Ito.SimpleIntegrand W hWm P ι) is a finite family of pieces
φᵢ 1_{(aᵢ, bᵢ]}(s) with 0 ≤ aᵢ ≤ bᵢ, each value φᵢ measurable with respect to the natural
filtration Gardiner.Ito.wienerFiltration at its left endpoint aᵢ (nonanticipation) and square
integrable. The pieces may overlap. Its process is G(s) = Σᵢ φᵢ 1_{(aᵢ,bᵢ]}(s)
(SimpleIntegrand.process) and its stochastic integral is the left-point sum
∫ G dW = Σᵢ φᵢ (W(bᵢ) - W(aᵢ)) (SimpleIntegrand.integral); the running integral is
Σᵢ φᵢ (W(bᵢ ∧ t) - W(aᵢ ∧ t)) (SimpleIntegrand.running).
- Linearity:
SimpleIntegrand.integral_add,SimpleIntegrand.integral_smul(and the same forprocess,running). - Mean value (4.2.40)–(4.2.41):
⟨∫ G dW⟩ = 0(SimpleIntegrand.integral_mean). - Correlation formula / Itô isometry (4.2.42)–(4.2.44): for two simple integrands,
⟨∫ G dW ∫ H dW⟩ = Σᵢⱼ ⟨φᵢψⱼ⟩ |(aᵢ,bᵢ] ∩ (cⱼ,dⱼ]|(SimpleIntegrand.integral_mul_eq_sum)= ∫ ds ⟨G(s)H(s)⟩(SimpleIntegrand.integral_mul), in particular⟨(∫ G dW)²⟩ = ∫ ds ⟨G(s)²⟩(SimpleIntegrand.integral_sq). - Martingale property of the running integral with respect to the natural filtration
(
SimpleIntegrand.martingale_running), andrunning T = integralonceT ≥ bᵢ(SimpleIntegrand.running_eq_integral).
Instances for Gardiner.Ito.stdWiener are in SHSLib.Gardiner.Ito.WienerInstances. The
library's SimpleProcess is not used.
The overlap length |(a, b] ∩ (c, d]|.
Instances For
A random variable measurable w.r.t. the past 𝓕 s is independent of any function of a
future increment W t - W s.
E[Z (W t - W s)] = 0 for an integrable 𝓕 s-measurable Z.
E[Z (W t - W s)²] = E[Z] (t - s) for an integrable 𝓕 s-measurable Z.
Z (W t - W s) is square integrable for a square-integrable 𝓕 s-measurable Z.
The basic correlation identity. For square-integrable φ (measurable at time a) and
ψ (measurable at time c), E[φ ΔW_{(a,b]} · ψ ΔW_{(c,d]}] = E[φψ] |(a,b] ∩ (c,d]|.
E[φ (W v - W u) | 𝓕 u] = 0 for square-integrable 𝓕 u-measurable φ.
The indicator 1_{(a,b]} as a real function.
Equations
- Gardiner.Ito.ind a b s = (Set.Ioc a b).indicator (fun (x : ℝ) => 1) s
Instances For
A simple (elementary) nonanticipating integrand (Gardiner §4.2.1, §4.2.4):
G(s) = Σᵢ φᵢ 1_{(aᵢ, bᵢ]}(s), a finite sum over an index type ι, where 0 ≤ aᵢ ≤ bᵢ and
each value φᵢ is square integrable and measurable with respect to the natural filtration at the
left endpoint aᵢ. The pieces may overlap; a step function on a partition is the special case
of adjacent intervals.
- left : ι → ℝ
left endpoints
- right : ι → ℝ
right endpoints
- val : ι → Ω → ℝ
values
- adapted (i : ι) : MeasureTheory.StronglyMeasurable (self.val i)
- memLp (i : ι) : MeasureTheory.MemLp (self.val i) 2 P
Instances For
The integrand as a process, G(s, ω) = Σᵢ φᵢ(ω) 1_{(aᵢ, bᵢ]}(s).
Instances For
The stochastic integral of a simple integrand, ∫ G dW = Σᵢ φᵢ (W(bᵢ) - W(aᵢ))
(the left-point sum of Gardiner (4.2.41)).
Instances For
The running integral ∫₀ᵗ G dW = Σᵢ φᵢ (W(bᵢ ∧ t) - W(aᵢ ∧ t)).
Instances For
The sum of two simple integrands (concatenation of their pieces).
Equations
Instances For
A scalar multiple of a simple integrand.
Equations
Instances For
Linearity: ∫ (G + H) dW = ∫ G dW + ∫ H dW.
Linearity: ∫ cG dW = c ∫ G dW.
The simple integral is square integrable.
(4.2.40)–(4.2.41): the mean of the simple integral vanishes, ⟨∫ G dW⟩ = 0.
The correlation of two simple integrals as a double sum (Gardiner (4.2.43)–(4.2.44)):
⟨∫ G dW ∫ H dW⟩ = Σᵢ Σⱼ ⟨φᵢψⱼ⟩ |(aᵢ,bᵢ] ∩ (cⱼ,dⱼ]|.
(4.2.42): the correlation formula (Itô isometry) for simple integrands,
⟨∫ G dW ∫ H dW⟩ = ∫ ds ⟨G(s) H(s)⟩.
The Itô isometry for simple integrands, ⟨(∫ G dW)²⟩ = ∫ ds ⟨G(s)²⟩.
The running integral is a martingale #
The martingale property of the running simple integral: t ↦ ∫₀ᵗ G dW is a
martingale with respect to the natural filtration of W.
At any time after the last right endpoint the running integral is the full integral.