Using (A.10) and (A.11-4), we have the following inequality;
Pr[∖Thml( J - JT)| > δ I sup∖Q,(x)-Qθ(x)≤C}
≤ Pr{ max { ∖Thm2(J -J„,)| , ∖TW'∖Jt - Jtl)| } > δ I sup∖Qβ(x)-Q(x)∖≤CT}
, for all δ > 0 . (A.12)
Invoking Lemma 1 and condition A2(iii), we have
Pr{ sup∖ Qθ(x)-Qθ(x)∖≤Ct} → 1 as T → ∞ . (A.13)
By (A.8) and (A.9), as T → ∞, we have
Pr{ max { ∖Thm/2(Jt - Jtu)∖ , ∖Thm/2(Jt - Jtl)∖ } > δ } → 0, for all δ > 0.
(A.14)
Therefore, as T → ∞,
the L.H.S. of the inequality (A.12) - Pr {∖ Thm/2(Jt - Jt )∖ > δ } → 0 and
the L.H.S. of the inequality (A.12) → 0 .
In summary, we have that if both Thm/2(JT-JTU) =op(1) and Thm/2(JT-JTU) =op(1),
then Thm/2(^Jt - Jt ) = op (1). □
Step 3: Asymptotic equivalence.
In the remaining proof, we focus on showing that Thm/2(JT-JTU) =op(1), with the proof
of Thm/2(JT-JTL) =op(1) being treated similarly. The proof of Step 3 is close in lines
with the proof in Zheng (1998). Denote
HT(s,t,g)≡Kts{1(yt≤g(xt))-,}{1(ys≤g(xs))-,} and (A.15)
J[g]≡
1
T (T -1) hm
T
∑
t=1
T
∑HT(s,t,g).
s≠t
(A.16)
Then we have JT ≡ J[Q,] and JTU ≡ J[Q, -CT]. We decompose HT (s,t, g) into
three parts;
HT(s,t,g)=Kts{1(yt≤g(xt))-F(g(xt)∖zt)}{1(ys≤g(xs))-F(g(xs)∖zs)}
13
More intriguing information
1. Regional dynamics in mountain areas and the need for integrated policies2. Evolution of cognitive function via redeployment of brain areas
3. Inflation Targeting and Nonlinear Policy Rules: The Case of Asymmetric Preferences (new title: The Fed's monetary policy rule and U.S. inflation: The case of asymmetric preferences)
4. The name is absent
5. Spectral density bandwith choice and prewightening in the estimation of heteroskadasticity and autocorrelation consistent covariance matrices in panel data models
6. Ex post analysis of the regional impacts of major infrastructure: the Channel Tunnel 10 years on.
7. The name is absent
8. Revisiting The Bell Curve Debate Regarding the Effects of Cognitive Ability on Wages
9. The name is absent
10. Knowledge and Learning in Complex Urban Renewal Projects; Towards a Process Design