Spectral calibration of exponential Levy models ??
23
-2δ2 ∞ (1
-∞
- e-v2)2U-2v-2U dv
ε-2δ2U-1 - ε-2δ(2s+5)/(s+2).
Thus, the Kullback-Leibler divergence remains small for δ — ε(2s+4)/(2s+5)
with a small constant, which gives the asymptotic lower bound for γ .
For the lower bound of λ we perturb the triplet T0 leaving γ0 and σ0 = 0
fixed and putting
Fμ 1(u) := Fμ0(u) + δe-u(u-i)/U2.
By similar estimates as for γ, when choosing U — δ-1 / (s+1) with a suffi-
ciently small constant, the perturbation T1 lies in Gs (R, 0) and the Kullback-
Leibler divergence is asymptotically bounded by
KL(Tι∣T)) . ε-2δ2U-3 — ε-2δ(2s+5)/(s + 1).
The basic lower bound results yields the asserted lower bound for λ.
7.3. Lower bound for μ in the case σ > 0
The interesting deviation from standard proofs of lower bounds (see e.g.
Butucea and Matias (2005)) for severely ill-posed problems is that we face
the restriction that Fμ is analytic in a strip parallel to the real line and is
uniquely identifiable from its values on any open set. So, let T0 = (σ2,γ0, μ0)
with σ0 > 0 be a Levy triplet from the interior of Gs (R,σmax ). Consider
the perturbation T1 = (σ2 ,γ0 ,μ 1) with
Fμ 1(u) := Fμ0(u) + δm 1 /4e-(Tσ2u2/m)m/2(Tσ2∣m)mum(u - i)m, u ∈ R .
for m ∈ N, δ > 0. Then we have uniformly for m → ∞ and δ → 0
∣∣μ 1 - μо k2 2 = p2∏τ=∣ fo e-vv(1+2m)/2m (1 + m- 1 v- 1 /m)m dv — δ2 ■
Similarly, for k = 1, . . . , s we derive uniformly in m and δ
∣∣μ(k ) - μ0k)∣∣l2 = √2∏∣∣ukF(μ 1 - μθ)(u)∣∣l2 — δmk/2,
∣∣μ 1s) - μ0s)∣∣∞ 6 ∣∣usF(μ 1 - μо)(u)∣∣li 6 δms/2- 1 /4.
Therefore choosing δ — m-s/2 with a small constant yields T1 ∈ Gs (R, σmax)
because we then also have that μ 1 is real-valued and T1 satisfies the mar-
tingale condition and Assumption 1.
By the same arguments as before and by Stirling’s formula to estimate
the Gamma function, the Kullback-Leibler divergence between the observa-
tions under T0 and under T1 is asymptotically bounded by
4ε
-2
∞
I1P о ,t ( u
- i ) 12 T2 IF ( μ 1
- μ0)(u)12(u4 + u2) 1 du