Exact asymptotic behavior of VarΠ

The proof of Theorem 2.1 is organized as follows. First Subsection 3.1, we determine the exact asymptotic behavior of the variance of Π n ǫ n . Then Subsection 3.2, we prove a central limit theorem for Π n ǫ n . By means of a de-Poissonization result Subsection 3.3, we then infer 3.3. In a final step Subsection 3.4 we prove 3.4, which completes the proof of Theorem 2.1. This Poissonizationde-Poissonization methodology goes back to at least Beirlant, Györfi, and Lugosi 1994.

3.1 Exact asymptotic behavior of VarΠ

n ǫ n Let ∆ n x = 1{π n x 0} − 1{ f x 0} . In the sequel, the letter C will denote a positive constant, the value of which may vary from line to line. Let ǫ n be the sequence of positive real numbers defined in 3.1. In this subsection, we intend to prove that, under the conditions of Theorem 2.1, lim n →∞ r n r d n Var Π n ǫ n = σ 2 f , 3.5 where σ 2 f is as in 2.3. Towards this goal, observe first that Π n ǫ n = Z ˜ E n 1{π n x 0} − 1{ f x 0} d x , where we set ˜ E n = E n ∩ S r n f , with S r n f = ¦ x ∈ R d : distx, S f ≤ r n © . Clearly, VarΠ n ǫ n = Z ˜ E n Z ˜ E n C ∆ n x, ∆ n y dxd y, where here and elsewhere C denotes ‘covariance’. Since ∆ n x and ∆ n y are independent when- ever kx − yk 2r n , we may write Var Π n ǫ n = Z ˜ E n Z ˜ E n 1 kx − yk ≤ 2r n C ∆ n x, ∆ n y dxd y. Using the change of variable y = x + 2r n u, we obtain VarΠ n ǫ n 2624 = 2 d r d n Z R d Z R d 1 ˜ E n x1 ˜ E n x + 2r n u 1 B0,1 uC ∆ n x, ∆ n x + 2r n u dxdu. By construction, whenever n is large enough, ˜ E n is included in the tubular neighborhood V ∂ S f , ρ of ∂ S f of radius ρ 0. In this case, each x ∈ ˜ E n may be written as x = p + ve p as described in 2.1. Hence, for all large enough n, we obtain VarΠ n ǫ n = 2 d r d n Z ∂ S f Z ρ −r n Z B0,1 1 ˜ E n p + ve p 1 ˜ E n p + ve p + 2r n u × Θp, vC € ∆ n p + ve p , ∆ n p + ve p + 2r n u Š dudv v σ dp. For all p ∈ ∂ S f , let κ p ǫ n be the distance between p and the point x of the set {x ∈ R d : f x = ǫ n } such that the vector x − p is orthogonal to ∂ S f . Using the change of variable v = t p nr d n , we may write Var Π n ǫ n = 2 d r d n p nr d n Z ∂ S f Z p nr d n κ p ǫ n − p nr d+2 n Z B0,1 1 ˜ E n   p + t p nr d n e p + 2r n u   Θ   p, t p nr d n   × C   ∆ n   p + t p nr d n e p   , ∆ n   p + t p nr d n e p + 2r n u     dudt v σ dp. For a justification of this change of variable, refer to equation 2.2 and equation 4.2 in the Ap- pendix. By conditions r.i and r.iii, nr d+2 n → 0. Consequently, r n r d n Var Π n ǫ n = o1 + 2 d Z ∂ S f Z p nr d n κ p ǫ n Z B0,1 1 ˜ E n   p + t p nr d n se p + 2r n u   Θ   p, t p nr d n   × C   ∆ n   p + t p nr d n e p   , ∆ n   p + t p nr d n e p + 2r n u     dudt v σ dp. 3.6 To get the limit as n → ∞ of the above integral, we will need the following lemma, whose proof is deferred to the end of the subsection. Lemma 3.1. Let p ∈ ∂ S f , t 0 and u ∈ B0, 1 be fixed. Suppose that the conditions of Theorem 2.1 hold. Then lim n →∞ C   ∆ n   p + t p nr d n e p   , ∆ n   p + t p nr d n e p + 2r n u     = Φp, t, u, where Φp, t, u is defined in Theorem 2.1. 2625 Returning to the proof of 3.5, we notice that by 4.4 in the Appendix and e.ii we have p nr d n κ p ǫ n → ∞ as n → ∞ and Θp, 0 = 1. Therefore, using Lemma 3.1 and the fact that for all t 0 and u ∈ B0, 1 1 ˜ E n   p + t p nr d n + 2r n u   → 1 as n → ∞, we conclude that the function inside the integral in 3.6 converges pointwise to Φp, t, u as n → ∞. We now proceed to sufficiently bound the function inside the integral in 3.6 to be able to apply the Lebesgue dominated convergence theorem. Towards this goal, fix p ∈ ∂ S f , u ∈ B0, 1 and t ≤ p nr d n κ p ǫ n . Since ∆ n x ≤ 1 for all x ∈ R d , using the inequality |CY 1 , Y 2 | ≤ 2E|Y 1 | whenever |Y 2 | ≤ 1, we have C   ∆ n   p + t p nr d n e p   , ∆ n   p + t p nr d n e p + 2r n u     ≤ 2 E ∆ n   p + t p nr d n e p   . 3.7 By the bound in 4.3 in the Appendix, we see that sup p ∈∂ S f κ p ǫ n → 0 as n → ∞. 3.8 Then, since e p is a normal vector to ∂ S f at p which is directed towards the interior of S f , there exists an integer N independent of p, t and u such that, for all n ≥ N , the point p + t p nr d n e p belongs to the interior of S f . Therefore, f p + t p nr d n e p 0 and, letting ϕ n x = P X ∈ Bx, r n , we obtain E ∆ n   p + t p nr d n e p   = P   π n   p + t p nr d n e p   = 0   = E  P   ∀i ≤ N n : X i ∈ B   p + t p nr d n e p , r n   N n     = E   1 − ϕ n   p + t p nr d n e p     N n = exp   −nϕ n   p + t p nr d n e p     , 3.9 where we used the fact that N n is a mean n Poisson distributed random variable independent of the sample. By a slight adaptation of the proof of Lemma A.1 in Biau, Cadre, and Pelletier 2008 and 2626 under Assumption 1-b, one deduces that for all x ∈ V ∂ S f , ρ ∩ ◦ S f , there exists a quantity K n x such that ϕ n x = r d n ω d f x + r d+2 n K n x and sup n sup x ∈V ∂ S f , ρ∩ ◦ S f |K n x| ∞. 3.10 For all x in V ∂ S f , ρ written as x = p + ue p with p ∈ ∂ S f and 0 ≤ u ≤ ρ, a Taylor expansion of f at p gives the expression f x = 1 2 D 2 e p f p + ξe p u 2 , for some 0 ≤ ξ ≤ u since, by Assumption 1-b, D e p f p = 0. Thus, in our context, expanding f at p, we may write n ϕ n   p + t p nr d n e p   = ω d D 2 e p f p + ξe p t 2 2 + nr d+2 n R n p, t, for some 0 ≤ ξ ≤ κ p ǫ n , and where R n p, t satisfies sup n sup § R n p, t : p ∈ ∂ S f and 0 ≤ t ≤ Æ nr d n κ p ǫ n ª ∞. Furthermore, by 3.8, each point p + ξe p falls in the tubular neighborhood V ∂ S f , ρ for all large enough n. Consequently, by Assumption 2-c there exists α 0 independent of n and N 1 ≥ N independent of p, t and u such that, for all n ≥ N 1 , inf p ∈∂ S f D 2 e p f p + ξe p 2α. This, together with identity 3.9 and r.i, r.iii, which imply nr d+2 n → 0, leads to E ∆ n   p + t p nr d n e p   ≤ C exp−ω d αt 2 3.11 for n ≥ N 1 and all ≤ t ≤ Æ nr d n sup p ∈∂ S f κ p ǫ n . Thus, using inequality 3.11, we deduce that the function on the left hand side of 3.7 is dominated by an integrable function of p, t, u, which is independent of n provided n ≥ N 1 . Finally, we are in a position to apply the Lebesgue dominated convergence theorem, to conclude that lim n →∞ r n r d n Var Π n ǫ n = 2 d Z ∂ S f Z ∞ Z B0,1 Φp, t, ududt v σ dp = σ 2 f . To be complete, it remains to prove Lemma 3.1. Proof of Lemma 3.1 Let x n = p + t p nr d n e p . Since nr d n → ∞ and nr d+2 n → 0, both x n and x n + 2r n u lie in the interior of S f for all large enough n. As a consequence, f x n 0 and f x n + 2r n u 0 for all large enough n. Thus, C ∆ n x n , ∆ n x n + 2r n u 2627 = C 1{π n x n = 0}, 1{π n x n + 2r n u = 0 } = P π n x n = 0, π n x n + 2r n u = 0 − P π n x n = 0 P π n x n + 2r n u = 0 = P ∀i ≤ N n : X i ∈ Bx n , r n ∪ Bx n + 2r n u, r n − P ∀i ≤ N n : X i ∈ Bx n , r n P ∀i ≤ N n : X i ∈ Bx n + 2r n u, r n = exp −nµ Bx n , r n ∪ Bx n + 2r n u, r n − exp −nµBx n , r n − nµBx n + 2r n u, r n , where µ denotes the distribution of X . Let B n = Bx n , r n ∩ Bx n + 2r n u, r n . Using the equality µ Bx n , r n ∪ Bx n + 2r n u, r n = ϕ n x n + ϕ n x n + 2r n u − µB n , we obtain C ∆ n x n , ∆ n x n + 2r n u 3.12 = exp −n ϕ n x n + ϕ n x n + 2r n u exp nµB n − 1 . Now, µB n may be expressed as µB n = f x n λB n + Z B n f v − f x n dv. Since f is of class C 1 on R d , by developing f at x n in the above integral, we obtain Z B n f v − f x n dv = r d+1 n R n , where R n satisfies R n ≤ C sup K kgrad f k, and K is some compact subset of R d containing ∂ S f and of nonempty interior. Next, note that λB n = r d n βu, where βu = λ B0, 1 ∩ B2u, 1 . Therefore, expanding f at p in the direction e p , we obtain µB n = βu t 2 2n D 2 e p f p + ξ t p nr d n e p + r d+1 n R n , where ξ ∈ 0, 1. Hence by r.iii, lim n →∞ n µB n = βuD 2 e p f p t 2 2 . The above limit, together with identity 3.12 and 3.10, leads to the desired result. ƒ 2628

3.2 Central limit theorem for Π

Dokumen yang terkait

AN ALIS IS YU RID IS PUT USAN BE B AS DAL AM P E RKAR A TIND AK P IDA NA P E NY E RTA AN M E L AK U K A N P R AK T IK K E DO K T E RA N YA NG M E N G A K IB ATK AN M ATINYA P AS IE N ( PUT USA N N O MOR: 9 0/PID.B /2011/ PN.MD O)

0 82 16

ANALISIS FAKTOR YANGMEMPENGARUHI FERTILITAS PASANGAN USIA SUBUR DI DESA SEMBORO KECAMATAN SEMBORO KABUPATEN JEMBER TAHUN 2011

2 53 20

EFEKTIVITAS PENDIDIKAN KESEHATAN TENTANG PERTOLONGAN PERTAMA PADA KECELAKAAN (P3K) TERHADAP SIKAP MASYARAKAT DALAM PENANGANAN KORBAN KECELAKAAN LALU LINTAS (Studi Di Wilayah RT 05 RW 04 Kelurahan Sukun Kota Malang)

45 393 31

FAKTOR – FAKTOR YANG MEMPENGARUHI PENYERAPAN TENAGA KERJA INDUSTRI PENGOLAHAN BESAR DAN MENENGAH PADA TINGKAT KABUPATEN / KOTA DI JAWA TIMUR TAHUN 2006 - 2011

1 35 26

A DISCOURSE ANALYSIS ON “SPA: REGAIN BALANCE OF YOUR INNER AND OUTER BEAUTY” IN THE JAKARTA POST ON 4 MARCH 2011

9 161 13

Pengaruh kualitas aktiva produktif dan non performing financing terhadap return on asset perbankan syariah (Studi Pada 3 Bank Umum Syariah Tahun 2011 – 2014)

6 101 0

Pengaruh pemahaman fiqh muamalat mahasiswa terhadap keputusan membeli produk fashion palsu (study pada mahasiswa angkatan 2011 & 2012 prodi muamalat fakultas syariah dan hukum UIN Syarif Hidayatullah Jakarta)

0 22 0

Pendidikan Agama Islam Untuk Kelas 3 SD Kelas 3 Suyanto Suyoto 2011

4 108 178

ANALISIS NOTA KESEPAHAMAN ANTARA BANK INDONESIA, POLRI, DAN KEJAKSAAN REPUBLIK INDONESIA TAHUN 2011 SEBAGAI MEKANISME PERCEPATAN PENANGANAN TINDAK PIDANA PERBANKAN KHUSUSNYA BANK INDONESIA SEBAGAI PIHAK PELAPOR

1 17 40

KOORDINASI OTORITAS JASA KEUANGAN (OJK) DENGAN LEMBAGA PENJAMIN SIMPANAN (LPS) DAN BANK INDONESIA (BI) DALAM UPAYA PENANGANAN BANK BERMASALAH BERDASARKAN UNDANG-UNDANG RI NOMOR 21 TAHUN 2011 TENTANG OTORITAS JASA KEUANGAN

3 32 52