The Stable Limit Theorem

Size: px
Start display at page:

Download "The Stable Limit Theorem"

Transcription

1 The Stable Limit Theorem Thomas oy September 6, 0 Itroductio Stable distributios are a class of probability distributios that have may useful mathematical properties such as skewess ad heavy tails. They have bee used as a model for differet types of physical ad ecoomic systems. I practice it has bee difficult to use stable distributios due to the fact that the desities ad distributio fuctios have closed formulas oly i specific cases (Gaussia, Cauchy ad Lévy). However, thaks to recet reliable computer programs, it is possible to compute stable desities, distributio fuctios ad quatiles. The focus of this paper will be the Stable Limit Theorem. I the course of this paper, we will try to rewrite the proof writte by William Feller [] ad add certai details that were omitted i his book. Afterwards, we will give some examples where the Stable Limit Theorem ca be applied. Stable Distributios Defiitio. (Defiitio VI.3. of []). A distributio F is ifiitely divisible if for each it ca be represeted as the distributio of the sum S = X, X, of idepedet radom variables with commo distributio F. We deote by ϕ Y the characteristic fuctio of a radom variable Y, i.e. ϕ Y (u) = E(e iuy ), where u. The ext result gives the represetatio of the characteristic fuctio of a ifiitely divisible radom variable usig a itegral with respect to a caoical measure. (See Defiitio A., Appedix A, for the defiitio of a caoical measure.) Theorem. (Theorem XVII.. of []). The class of characteristic fuctios of ifiitely divisible distributios coicides with the class of fuctios of the form: { ϕ Y (u) = E(e iuy ) = exp iub + e iux } iu si x x M(dx) Project supported by a NSEC Udergraduate esearch Award at Uiversity of Ottawa, durig the summer of 0. (Supervisor: aluca Bala) ()

2 where M is a caoical measure ad b is a real umber. I this case, we write X ID F eller (b, M). emark.3. I the right-had side of (), the itegrad is defied to be u whe x = 0. Hece, if M = Cδ 0, where δ 0 deotes the Dirac measure at 0, the Y has ormal N(b, C) distributio. emark.4. There is a alterative way of writig the characteristic fuctio of a ifiitely divisible distributio as: ϕ Y (u) = exp {iua σ u } + (e iux iux { x } )ν(dx) where a, σ 0 ad ν is a Lévy measure o, i.e. ν satisfies: (x )ν(dx) < ad ν({0}) = 0. This is called the Lévy represetatio. Here a b deotes the miimum betwee a ad b. I this case, we write Y ID(a, σ, ν). Note that if Y ID(a, σ, ν) the Y ID F eller (b, M) where b = a + (si x x { x } )ν(dx) ad M(dx) = x ν(dx) + σ δ 0 (dx) Defiitio.5 (Defiitio VI.. of []). Let F be the law of a radom variable X, (X i ) i be i.i.d. copies of X ad S = = X i. The distributio F is stable if for each there exist costats c > 0, d such that ad F is ot cocetrated at oe poit. S d = c X + d () Theorem.6 (Theorem VI.. of []). The ormig costats i () are of the form c = /α with 0 < α. Defiitio.7. A radom variable Y with characteristic futio ϕ Y (u) = e ψ Y (u), u, where ψ Y (u) is give by ( iuδ u α γ α isg(u)β ta πα ), α ψ Y (u) = iuδ u γ( + isg(u)β π l u ), α = (3) for some α (0, ], γ > 0, β [, ] ad δ is said to have a α-stable S α (γ, β, δ) distributio. I this case, we write Y S α (γ, β, δ). Usig Example XVII.3.(g)-(h) of [], we obtai the followig lemma.

3 Lemma.8. Assume that Y ID F eller (0, M α ), i.e. its characteristic fuctio is give by () where 0 < α < ad M α (dx) = [c + x α+ (0, ) (x) + c ( x) α+ (,0) (x)]dx (4) for some c +, c 0. The Y S α (γ, β, δ) where the parameters (γ, β, δ) are give by: Γ(3 α) (c γ α + + c ) = α( α)( α) cos πα, α (c + + c ) π (5), α = β = c + c c + + c (6) si x (c + c ) dx, 0 < α < 0 xα+ δ = x si x (c + c ) 0 x dx, < α < 0, α = (7) If Y ID F eller (b, M α ) where b ad M α is give by (4), the Y S α (γ, β, δ ) where δ = b + δ ad (γ, β, δ) are the same as above. emark.9. The costat δ i (7) ca also be expressed as: si x x M α(dx), 0 < α < δ = x si x x M α (dx), < α < 0, α = (8) To see this, ote that if α (0, ) δ = (c + 0 = (c + 0 = si x x x α+ dx c si x x x α+ dx + c si x x M α(dx) The cases α (, ) ad α = are similar. 0 0 ) si x x x α+ dx ) si x x x α+ dx Lemma.0. Assume that Y ID(0, 0, ν α ), where the measure ν α is give by: ν α (dx) = [c + x α (0, ) (x) + c ( x) α (,0) (x)]dx (9) 3

4 for some c +, c 0 ad α (0, ). The Y S α (γ, β, δ ) where (γ, β) are give by (5) ad (6), ad x { x } ν α (dx) 0 < α < δ = x { x } ν α (dx), < α < (0) (si x x { x } )ν α (dx), α = Proof. By emark.4, Y ID F eller (b, M α ), where M = M α (M α is give by (4)) ad b = (si x x { x } )ν α (dx) By Lemma.8, Y S α (γ, β, δ ), where (γ, β) are give by (5) ad (6), δ = b+δ ad δ is give by (8) (see emark.9). Hece, whe α < δ = (si x x { x } )ν α (dx) si x ν α (dx) = x { x } ν α (dx) The desired expressio for δ i the case α (, ) or α = follows similarly. 3 Domai of attractio Defiitio 3.. Let (X i ) i be idepedet idetically distributed (i.i.d.) radom variables with law F ad S = i= X i. Let Y be a radom variable with law G. We say that X belogs to the domai of attractio of Y if there exist > 0 ad b such that: emark 3.. Note that () is equivalet to S b d Y () (ϕ (u) ) ψ Y (u) for ay u () where ϕ is the characteristic fuctio of F, the probability distributio of X b, i.e. ϕ (u) = ϕ X a b(u) = ϕ X ( u ) e iub ad ϕ Y (u) = e ψ Y (u). To see this, ote that by the cotiuity of characteristic fuctios, () is equivalet to: ϕ S a b (u) ϕ Y (u) for ay u. (3) 4

5 Note that the characteristic fuctio of S b is give by: { [ ( )]} S E exp iu b ( ) = E exp Xj iu b a j= )]} (iu Xa b = { E [ exp = [ϕ (u)]. The (3) is equivalet to: ( [ϕ (u)] = + (ϕ ) (u) ) ϕ Y (u) = e ψ Y (u) for ay u which is i tur equivalet to (). We cosider the fuctio: U(x) = E ( X ( X x) ), x > 0. (4) Theorem 3.3. Let X, (X i ) i be i.i.d. radom variables with commo distributio F ad let S = i= X i. Suppose that there exist > 0 ad b such that () holds. The Y has a stable distributio with characteristic fuctio () where [ ( )] X b = lim β, β = E si b M = Cδ 0 or M = M α ad M α is give by (4). Proof. We deote We cosider separately two cases: M (dx) = x F (dx) (5) Case : Assume that F is symmetric, (i.e. P (X < x) = P (X > x) for ay x > 0) ad () holds with b = 0. I this case, F is the probability distributio of X. Note that, if F is symmetric, we cosider (X i ) i, a idepedet copy of (X i ) i ad we let X i = X i X i. The ( X i ) i is a sequece of i.i.d. radom variables with a symmetric distributio ad S = X i = S S i= 5

6 where S = i= X i. By Theorem.8 of [] ( S S ) b, b d (Y, Y ) where Y is a idepedet copy of Y. By applyig the Cotiuous Mappig Theorem of [] to the map of h :, h(x, y) = x y, we get: ( ) ( ) S S S = b b d Y Y = Ỹ. Hece, the symmetrized sequece ( X i ) i satisfies () with b = 0 ad the same as the origial sequece (X i ) i. Usig Corollary A.4, with C = ad b = 0, we get, by () that there exist a caoical measure M ad b [ such ( that )] M M properly ad X β b, where β = si xf (dx) = E si. I this case ψ Y (u) = iub + + e iux iu si x x M(dx). We ow prove various properties of the sequece ( ), which are cosequeces of the fact that S d Y. Sice M M properly, we have (accordig to Defiitio A..(i)) M ([ x, x]) M([ x, x]) for ay cotiuity poit x of M. Usig the defiitio (5) of M ad the fact that i this case F is the distributio of X, we have: M ([ x, x]) = [ x,x] (y)y F (dy) = ( ) X E { X = a a E ( X { X ax}). a We obtai that ( ) must satisfy: U( x) M([ x, x]) for ay cotiuity poit x of M. (6) a From Defiitio A..(ii), we get: M + (x) = x y M (dy) M + (x) = 6 x y (dy)

7 for ay cotiuity poit x of M. Note that: x y M (dy) = x y y F (dy) = (x, ) (y)f (dy) ( )) ( ) X = E ( (x, ) = P > x Xa We obtai: Similarly, = P (X > x) = ( F ( x)) [ F ( x)] M + (x) for ay cotiuity poit x of M. (7) F ( x) M ( x) for ay cotiuity poit x of M. Due to (), we have ϕ (u) = (ϕ (u) ) 0 i.e. ϕ (u) = ϕ X ( u ). This proves that ϕ X ( u ) = E (e i u X) a. Hece, Feller cocludes i [] (he details of Feller s reasoig are still ot clear to us), that u 0 i.e.. It remais to show the details of Feller s reasoig. We ow prove that To see this, ote that [ ( ϕ S (u) = E exp iu S )] = a Similarly, ϕ S + (u) = [ ϕ X ( u + j= + [ ( E exp iu X )] [ ( )] j u = ϕ X = [ϕ (u)] ϕ Y (u). )] = [ϕ +(u)] = [ϕ +(u)] + ϕ + (u) ϕ Y (u). The we get S d S d + Y ad Y. So, Feller cocludes that. + It looks ituitively right, but it remais to show the details of to prove this statemet. Usig (6) ad Lemma A.7 (Appedix A) with λ = a, χ(x) = M([ x, x]) ad D the set of cotiuity poits of M, we obtai that U is regularly varyig of idex ρ ad M([ x, x]) = Cx ρ, for some C > 0. Note that ρ 0. (If ρ < 0 7

8 the M() = lim x M([ x, x]) = lim x Cx ρ = 0, which is impossible). We deote ρ = α for some α. Hece for a slowly varyig fuctio l, ad U(x) = x α l(x) (8) M([ x, x]) = Cx α (9) If α =, the M([ x, x]) = C for ay x > 0 i.e. M = Cδ 0. If α <, the M is give by: ( M(dx) = C( α) x α+ (0, ) (x) + ) ( x) α+ (,0) (x) dx (0) because x x C( α) ( y α+ (0, ) (y) + ) ( y) α+ (,0) (y) dy = Cx α I this case, M has o atom at the origi, i.e. M({0}) = 0. Sice M is a caoical measure, M + (x) = Usig (0), we have M + (x) = We obtai that: x y C( α) y α+ dy = M + (x) = C y = = M([ x, x]). M(dy) <, for ay x > 0. y C( α) y α dy x C( α) x α. α α α x α, for ay x > 0. () Case : Assume ow that F is ot symmetric. Usig () ad Corollary A.4 with C = ad b = 0, we coclude that there exist a caoical measure M ad b such that M M properly ad β b, where I this case β = [ ( )] X si(x)f (dx) = E si b. + ψ Y (u) = iub + e iux iu si x x M(dx). 8

9 Sice M M properly, we have:. M ([ x, x]) M([ x, x]). M + (x) = x y M (dy) x y M(dy) = M + (x) for ay cotiuity poit x of M. We ote that, similarly to Case, x y M (dy) = [ F ( (x + b ))]. x 3. M x ( x) = y M (dy) y M(dy) = M ( x) for ay cotiuity poit x of M. Similarly, x y M (dy) = [F ( ( x + b ))]. Usig the defiitio (5) of M ad the fact that, i this case, F is the distributio of X b, we see that M ([ x, x]) = [ x,x] (y)y F (dy) = E [ ( ) ] X b { X b x} a Therefore from. we obtai that [ ( ) ] X E b { X b x} M([ x, x]) a which is the aalogue of (7) i the o-symmetric case. O the other had: M + (x) = x y y F (dy) = (x, ) (y)f (dy) ( )] ( ) X = E [ (x, ) b = P b > x Xa = P (X > (x + b )) = [ F ( (x + b )]. Similarly, M ( x) = F [ ( x + b )]. Usig. ad 3. we get: [ F ( (x + b ))] M + (x) ad F ( ( x + b )) M ( x) () for ay x > 0 such that x ad x are cotiuity poits of M. Note that b 0 To see this, recall that ϕ (u) (due to ()). But ϕ (u) = ϕ X ( u )e iub ad ϕ X ( u ) sice. (Note that we proved that i the case whe F is symmetric ad b = 0. I the geeral case, the symmetrized sequece ( X i ) i satisfies () with the same as (X i ) i. Hece by Case,. ) Hece e iub, i.e. b 0. 9

10 We claim that, due to the fact that b 0, we have: [ F ( (x + b )) + F ( x)] 0 ad [F ( ( x + b )) F ( x)] 0. It remais to show this claim is true. Hece, [ F ( x)] M + (x) ad [F ( x)] M ( x) (3) for ay x > 0 such that x ad x are cotiuity poits of M. This coclusio also eeds to be prove. We ow use Lemma A.7 three times:. Sice ( F )( x) M + (x) for ay x > 0 cotiuity poit of M, we have two situatios: either M + (x) = 0 for ay x or F is regularly varyig of idex α + ad M + (x) = A + x α+ for some A + > 0. Note that, because 0 = lim x M + (x) = A + lim x x α+, it is ecessary that α + > 0.. Sice ( F )( x) M ( x) for ay x > 0 such that x is a cotiuous poit of M, we have two situatios: either M ( x) = 0 for ay x or F is regularly varyig of idex α ad M ( x) = A x α for some A > 0. Note that, similarly, α > P ( X > x) M + (x) + M ( x) (4) for ay x > 0 such that x ad x are cotiuity poits of M. We deote V (x) = P ( X > x) (V is called the tail sum of X). We have two situatios: either M + (x) + M ( x) = 0 for ay x or V is regularly varyig of idex α ad M + (x) + M ( x) = Ax α for some A > 0. Note that, similarly, α > 0. If M + ad M do ot vaish idetically (i.e. some x), we get: M + (x) = M ( x) > 0 for A + x α+ + A x α = Ax α for ay x Hece α + = α = α (i.e. we have the same expoet α for both tails) ad A + + A = A. To summarize, we have two cases: either M + (x) = M ( x) = 0 for ay x (i.e. M = Cδ 0 ) or, M = M α where M α is give by (0) for some α > 0 ad c +, c > 0. Note that if M = M α, the α <. We get this by usig the fact that x M([0, x]) = c + y α+ dy < for ay x >

11 emark 3.4. Whe α (0, ), M α ([0, x]) = c + α x α ad M([ x, 0]) = c α. This gives us: Mα([ y, x]) = Cpx α + Cqy α (5) where C = c + + c α, p = c + c ad q =. (6) c + + c c + + c Whe α (0, ), we also have: M α + (x) = c + α x α = Cp α α x α ad Mα ( x) = c α x α = Cq α α x α. (7) elatios (5) ad (7) also hold whe α =, replacig M α by M = Cδ 0. Whe M = Cδ 0, the characteristic fuctio of Y is E[e iuy ] = exp{iub u C} i.e. Y has ormal distributio with mea b ad variace C. Whe M = M α, the log-characteristic fuctio of Y is give by: + e iuα iu si x ψ Y (u) = iub + x M α (dx), i.e. Y ID F eller (b, M α ) ( ) where b = lim β ad β = E si( X b ). I this case, accordig to Lemma.8, Y has S α (γ, β, δ ) distributio, with δ = δ + b. Theorem 3.5 (Theorem XVII.5. of []). (0) A distributio G posesses a domai of attractio if ad oly if it is stable. (i) The class of stable distributios coicides with the class of ifiitely divisible distributios with caoical measures give by M = Mδ 0 or M = M α, where M α is give by (4) for some c +, c 0, 0 < α <. (ii) The log-characteristic fuctio of a stable distributio is give by (3) for some (α, γ, β, δ) with α (0, ], γ > 0, β [, ], δ. (iii) If X has a stable distributio of idex α, the { x α P (X > x) Cp α α as x x α P (X < x) Cq α α as x where M([ y, x]) = Cpx α + Cqy α for some C > 0, p, q 0, p + q =. Proof. (0) By Theorem.6, if a radom variable Y possesses a domai of attractio, the Y has a stable distributio. Coversely, suppose that Y has a stable distributio. We show that Y belogs to its ow domai of attractio. Let (Y i ) i be i.i.d. (as Y ). Let S = Y Y.

12 By the defiitio of the stable distributio, S d = /α Y + d for a costat d. Hece S d /α d = Y d Y. (i) Suppose Y has a stable distributio. By (0), there exist i.i.d. radom variables (X i ) i such that S d b Y. By Theorem.6 Y IDF eller (b, M), with M = M α or M = Cδ 0. Coversely, suppose that Y ID F eller (b, M). By Lemma.8, Y has a logcharacteristic fuctio (3). (ii) We show that Y has a stable distributio if ad oly if Y has a logcharacteristic fuctio (3). Suppose Y has a stable distributio. The by (i), Y ID F eller (b, M) with M = M α or M = Cδ 0. By Lemma.8, Y has a log-characteristic fuctio (3). Suppose that Y has a log-characteristic fuctio (3). Let Y,...Y be i.i.d. copies of Y. We wat to prove that there exist c > 0 ad d such that: Y Y d = c Y + d By Theorem.6, if such a c exists, the c = /α. We have: [ ϕ Y+...+Y (u) = E e iu(y+...+y)] = [E(e iuy )] = [ϕ Y (u)] = e ψ Y (u) ad [ ϕ cx+d (u) = E e iu(cy +d)] = e iud E[e iucy ] = e iud ϕ Y (c u) = e iud+ψ Y (c u) It suffices to show that: ψ Y (u) = iud + ψ Y ( /α u) If α, we have by (3) : [ ( πα )] ψ Y (u) = iuδ u α γ α isg(u)β ta [ ( πα = iu( )δ + iuδ /α u α γ α isg( /α u)β ta )] with = iud + ψ Y ( /α u) The calculatio for α = is similar. d := iu( )δ (iii) Suppose that Y has a stable distributio. By (0), Y belogs to its ow S domai of attractio, i.e. b d /α Y, where S = Y Y ad (Y i ) i are i.i.d. copies of Y. By (3) ad (7), it follows that, for ay x > 0, [ F ( /α x)] M + (x) = Cp α α x α

13 ad [F ( /α x)] M (x) = Cq α α x α. Hece, for ay x > 0, ad Deote /α x = x. We get: x α P (Y > /α x) Cp α α ( /α x) α P (Y > /α x) Cp α α ( /α x) α P (Y < /α x) Cq α α. ad x α P (Y < /α x) Cq α α If lim x x α P (X > x) the lim x x α P (X > x) = lim x α P (X > x ) for ay sequece (x ), such that x. Take x = /α x. We still eed to see whe the limit exists ad why Feller makes these coclusios whe it does. 4 Coditios for X belogig to the domai of attractio of Stable(α) Lemma 4.. Let X be a arbitrary radom variable. For ay x > 0, defie U(x) = E ( X { X x} ) ad V (x) = P ( X > x). Suppose that E(X ) =, i.e. lim x U(x) =. (i) If U is regularly varyig with idex α for some 0 < α or V is regularly varyig with idex α, for some 0 < α, the x P ( X > x) lim = α x U(x) α. (8) (ii) Coversely, if (8) holds for some 0 < α <, the there exists a slowly varyig fuctio L 0 such that: U(x) αx α L 0 (x) ad P ( X > x) ( α)x α L 0 (x). If (8) holds for α =, the U is slowly varyig. Proof. Apply Theorem A.8 with a = ad b = 0. Theorem 4. (Theorem XVII.5. of []). Let X be a radom variable with law F ad U(x) = E ( X { X x} ). (i) If X belogs to some domai of attractio the: U(x) = x α l(x) (9) 3

14 for some 0 < α ad a slowly varyig fuctio l. (ii) Assume that X is o-degeerate ad U is slowly varyig (i.e. (9) holds for α = ). The X belogs to the domai of attractio of ormal distributio. (iii) Assume that (9) holds for some 0 < α <. The X belogs to the domai of attractio of a stable law of idex α if ad oly if the tails are balaced, i.e. lim x for some p, q 0 with p + q =. P (X > x) P (X < x) = p ad lim P ( X > x) x P ( X > x) = q (30) Proof. (i) By Defiitio 3., there exist ( ) ad (b ) such that S d b Y, where S = i= X i ad (X i ) i are i.i.d. copies of X. By Theorem 3.3, Y ID F eller (b, M) where M = M α or M = Cδ 0 ad M α is give by (4). Assume M = M α with 0 < α <. By (3) ad (7): P ( X > x) M + (x) + M ( x) = C α α x α. By Lemma A.7, it follows that V (x) = P ( X > x) is regularly varyig with idex α. By Lemma 4..(i), (8) holds. By Lemma 4..(ii), U is regularly varyig with idex α, i.e. (9) holds. Assume that M = Cδ 0. I this case, M + (x) = M ( x) = 0 for ay x > 0 ad Y has ormal distributio. By (3), P ( X > ) 0. (3) We ow prove that U is slowly varyig. Sice S d b Y, it follows that ( ) S does ot coverge i probability to 0. Moreover, ( Xk P ) > for some k 0. (3) This statemet eeds to be prove i details. We will prove that there exist C > 0 ad 0 such that U( ) ( ) X a = E k a { Xk } > C for ay 0, N (33) k= where N N is a subsequece. Usig (3) ad (33), it follows that Hece U( ) a P ( X > ) = U() P ( X > ). a a P ( X > ) U( ) 4 0

15 ad relatio (8) holds with α =. By Lemma 4..(ii), ( ) U is slowly varyig. It S remais to prove (33). To see this, ote that sice does ot coverge i probability to 0, there exists ɛ 0 > 0, C 0 > 0 ad a subsequece N N such that P ( S > ɛ 0 ) > C 0 for ay N. (34) Writig X k = X k { Xk } + X k { Xk >}, we obtai that S X k { Xk a } + X k { Xk >a } k= ad hece ( ) ( ) ( ) S X k P > ɛ 0 P { Xk a } > ɛ 0 X k +P { Xk >a } > ɛ 0. k= (35) By Chebyshev s iequality (ad assumig that X k has a symmetric distributio) ( ) X k P { Xk a } > ɛ 0 4 X k ɛ E k= 0 { Xk a } = 4 U( ) ɛ k= 0 a. (36) O the other had, for the secod term o the right had side of (35) we have { } X k { Xk >a } > ɛ 0 { X k > for some k } k= sice if X k < for all k the greater tha ɛ 0. Hece, by (3) ( ) lim P X k { Xk >a } > ɛ 0 k= k= k= k= X k { Xk >} = 0, which caot be lim P ( X k > for some k ) = 0. Usig (34), (35), (36) ad (37) it follows that ( ) S C 0 < lim, N P > ɛ 0 4 U( ) ɛ lim, N 0 a (37) Hece there exists a iteger 0 such that U() 0, N. This cocludes the proof of (33). a C 0 ɛ 0 4 =: C for ay (ii) This will follow from Theorem 4. (below). 5

16 (iii) Assume first that X belogs to the domai of attractio of a stable law of idex α <. By (3), (4) ad (7) { P (X > x) M + (x) = Cp α α x α ad P ( X > x) M + (x) + M ( x) = C α α x α. We the get: P (X > x) P ( X > x) M + (x) M + (x) + M =: p for ay x > 0 ( x) P (X > x) If lim x P ( X > x) exists, the lim x for ay x. We take x =. P (X > x) P ( X > x) = lim P (X > x ) P ( X > x ) Assume ext that (30) holds. The fact that X belogs to domai of attractio of a stable law will follow by Theorem 4. below. We still eed to kow whe the limit exists. Theorem 4. has two importat corollaries. Corollary 4.3. A o-degeerate radom variable X belogs to the domai of attractio of the ormal distributio (ad we write X DAN) if ad oly if U is slowly varyig, or equivaletly: x P ( X > x) lim = 0 (38) x U(x) Proof. We first show that X DAN if ad oly if U is slowly varyig. Assume X DAN. By Theorem 4..(i), U is slowly varyig. Assume that U is slowly varyig. By Theorem 4..(ii), X DAN. Note that, by Lemma 4., U is slowly varyig if ad oly if (38) holds. Corollary 4.4. A radom variable X belogs to the domai of attractio of a stable law of idex α, with 0 < α <, if ad oly if the followig coditios hold: (i) P ( X > x) x α L(x), for some slowly varyig fuctio L. P (X > x) (ii) P ( X > x) p ad P (X < x) q for some p, q 0, p + q =. P ( X > x) Coditio (i) is equivalet to (8). Proof. Assume first that X belogs to the domai of attractio of a stable law of idex α, with 0 < α <. The U(x) x α l(x) (by Theorem 4..(i)). By Lemma 4..(i), (8) holds. By Lemma 4..(ii), with αl 0 = l, P ( X > x) α α x α l(x). Hece coditio (i) holds with L = α α l. Coditio (ii) holds by Theorem 4..(iii). Assume (i) ad (ii) hold. By (i), V (x) = P ( X > x) is regularly varyig with idex α. By Lemma 4..(i), (8) holds. By Lemma 4..(ii), (9) holds. 6

17 By Theorem 4..(iii), (9) ad (ii) imply that X belogs to the domai of attractio of a stable law of idex α, with 0 < α <. The fact that (i) is equivalet to (8) follows by Lemma 4.. The followig result is kow i the literature as Karamata Theorem. Lemma 4.5. (i) Assume that U(x) = E ( X { X x} ) is regularly varyig of idex α, where 0 < α. The, for ay β < α, E( X β ) < ad x β lim x U(x) E ( X β ) α { X >x} = α β. (39) (ii) Assume that V (x) = P ( X > x) is regularly varyig of idex α, where 0 < α <. The, for ay β > α, E( X β ) = ad E ( ) X β { X x} lim x x β = α (40) P ( X > x) β α Proof. (i) elatio (39) follows usig Theorem A.8 with a = ad b = β, i.e. U(x) = U (x) = E ( ) X { X x} ad Vβ (x) = E ( ) X β { X >x}. Because E ( ) X β { X x} x β < ad E ( ) X β { X >x} <, we get: E( X β ) = E ( ) ( ) X β { X x} + E X β { X >x} < (ii) elatio (40) follows usig Theorem A.8 with a = β ad b = 0, i.e. U β (x) = E ( ) X β { X x} ad V0 (x) = P ( X > x). From (40), we get that: E ( ) X β α { X x} β α xβ P ( X > x). But lim E ( X β ) { X x} = E( X β α ) ad lim x x β α xβ P ( X > x) =. It follows that E( X β ) =. The followig result shows that if covergece () holds the the ormalizig sequece ( ) must satisfy a very importat coditio. Lemma 4.6. Suppose that () holds, where S = i= X i ad (X i ) i are i.i.d. radom variables, with commo distributio F. The the ormalizig sequece ( ) has to satisfy: for ay x > 0. I particular U( x) Cx α, (4) a U( ) C. (4) a Proof. I the proof of Theorem 3.3 the case whe F is arbitrary, we have see that ( ) has to satisfy: P ( X > x) M + (x) + M ( x) = C α α x α for some 0 < α <, (43) 7

18 (see relatios (3) ad (7)), or P ( X > x) 0. The measure M is give by M = M α, respectively M = Cδ 0, where M α is give by (4). To prove (4), i the case of α <, let V (x) = P ( X > x). By Lemma A.7 ad (43), V is regularly varyig of idex α. By Lemma 4..(i), (8) holds. This a gives us: x P ( X > x) α a for ay x > 0, i.e: P ( X > x) U( x) α U( x) α x. By multiplyig the omiator ad deomiator by, we get α P ( X > x) a U(a x) α α x. elatio (4) follows by (43). I the case α =, relatio (4) follows from (6) (assumig that F is symmetric) ad the fact that M = Cδ 0, ad hece M([ x, x]) = C. (We still do t kow why (4) holds i the o-symmetric case). The followig result gives a explicit recipe for costructig a sequece ( ) satisfyig (4). Lemma 4.7. Assume that is U is regularly varyig of idex α, with 0 < α. i.e. Let C > 0 arbitrary. Defie { } U(x) = if x > 0; x C where we use the covetio if =. The: (i) is fiite, (ii) ( ) satisfies (4), (iii) lim =. { U(x) Proof. (i) We have to prove that the set A = x > 0; x C } is o empty. To see this, say U(x) = x α l(x) for a slowly varyig fuctio l. We have to fid some x > 0 such that x U(x) = x x α l(x) = x α l(x) C. l(x) This follows sice lim x x α = 0. { U(x) (ii) Note that + sice x; x C } By the defiitio of, we get: we get U( ) a > C { x; U(x) x C }. + U(x) x > C for ay x <. Takig x =, for ay. We apply this result with +. We obtai: U( ) a > C +. (44) 8

19 O the other had, by the defiitio of : U( ) a C (45) By combiig (44) ad (45), we get: which is equivalet to C + < U() a C C + < U() elatio (4) follows from the fact that a C. +. (iii) Sice ( ) is o-decreasig sequece, lim =: C 0 [0, ] exists. Suppose C 0 <. By (4), we kow that U( ) Ca. Note that U( ) U(C 0 ) as. Hece U( ), which is a cotradictio. emark 4.8. Note that if U is regularly varyig of idex α, ad ( ) is chose to satisfy (4), the it also satisfies (4). To see this, ote that l(x) (because l is slowly varyig ad ) l( ) ad a α l( ) = a α l( ) a = U() a C. Hece a U( x) = a a α x α l( x) = (a α l( )) l(x) l( ) x α Cx α. The followig result specifies a more coveiet way of choosig the sequece ( ), i the case α (0, ). Lemma 4.9. Suppose that α (0, ) ad ( ) satisfies: P ( X > ) C (46) for some C > 0. If V (x) = P ( X > x) is regularly varyig of idex α (i.e. coditio (i) of Corollary 4.4 holds), the ( ) satisfies (4) with C = α C α. 9

20 Proof. By hypothesis, the fuctio V (x) = P ( X > x) is regularly varyig of idex α. By Lemma 4..(i) (8) holds. By Lemma 4..(ii) with L = ( α)l 0 we obtai: U(x) α α x α L(x), i.e. U is regularly varyig of idex α. By (8) a P ( X > ) U( ) U( ) a = P ( X > ) U() a α C = C, i.e. (4) holds. α α. Usig (46), we obtai: α emark 4.0. Assume α (0, ). A example of a sequece ( ) satisfyig (46) is: { = if x > 0; P ( X > x) C } { = if x > 0; P ( X x) C }. This ca be proved similarly to Lemma 4.7. We deote by F 0 the law of X. The { = if x > 0; F 0 (x) C } ( =: F0 C ), where F 0 deotes the geeralized iverse of the o-decreasig fuctio F 0. If F 0 is a cotiuous distributio, the this is equivalet to sayig that: F 0 ( ) = C, i.e. is the C -quatile of F 0. (ecall that z α is called the α-quatile of a radom variable Z if P (Z > z α ) = α, i,e, P (Z z α ) = α.) I the remaiig part of this sectio, we cocetrate o provig that relatio () holds, uder suitable coditio imposed o the uderlyig distributio of the sequece (X i ) i of i.i.d. radom variables. For this, we will assume that ( ) is o-decreasig sequece of costats with lim =, which satisfies (4). We begi with a prelemiary result. Propositio 4.. Let F be the law of X ad M (dx) = x F (dx). Let U be defied by (4). Assume that U is regularly varyig of idex α, 0 < α. Let ( ) be a sequece satisfyig (4), ad. If α <, we assume i additio that the tails of X are balaced, i.e. (30) holds. The M M properly where M = Cδ 0 if α =, ad if α (0, ), M = M α, where M α is give by (4) with costats c +, c satisfyig (6). Moreover, e iux iu si x e iux iu si x x M (dx) x M(dx). (47) 0

21 Proof. Note that M is caoical measure o because M (dx) = x F (dx) < ad x M (dx) = x x M is also a caoical measure. By emark 4.8, (4) holds. We claim that x x M ([ x, x]) M([ x, x]) for ay x > 0. (48) This follows from (4) ad the defiitio of M sice [ ] U( x) X a = E a { X ax} = [ x,x] (y)y F (dy) = M ([ x, x]) ad M([ x, x]) = Cx α. From (4), we also obtai that, for ay α (0, ], P ( X > x) M + (x) + M ( x) for ay x > 0 (49) where M + ad M are give by Defiitio A.. To see this, ote that sice U is regularly varyig of idex α, x P ( X > x) U(x) α α, as x by Lemma 4.. Sice x, it follows that for ay x > 0, which is equivalet to sayig that: a x P ( X > x) U( x) α α a U( x) P ( X > x) α α x. x x F (dx) <. By (4), we have: a U( x). We the obtai: Cx α P ( X > x) C α α x x α which is equivalet to (49), sice due to the defiito of M, M + (x) + M ( x) = C α α x α for ay x > 0. (50)

22 I particular, whe α =, M + (x) = M ( x) = 0 for all x > 0 (sice M = Cδ 0 ), ad relatio (49) becomes: P ( X > x) 0 for ay x > 0. Whe α <, the tails are balaced, (i.e. (30) holds) by hypothesis. Due to (49) ad (50), we obtai that P (X > x) Cp α α x α ad P (X < x) Cq α α x α (5) To see this, ote that P (X > x) = P (X > x) P ( X > x) P ( X > x) = Cp α α x α. The secod part of (5) is obtaied similarly. Due to the defiitio of F ad (7), (5) ca be writte as: M + (x) = x y y F (dy) x y M(dy) = M + (x) x M x ( x) = y y F (dy) y M(dy) = M ( x) Puttig together (48) ad (5) we get: (5) M ([ x, x]) M([ x, x]), M + (x) M + (x), M ( x) M ( x) i.e. M M properly (accordig to Defiitio A.) Defie ψ (u) = ψ(u) = e iux iu si x x M (dx) e iux iu si x x M(dx). Sice M M properly (i the sese of defiitio A.), by Theorem A.3, it follows that ψ (u) ψ(u). This proves (47). The followig result is called the Stable Limit Theorem, which is the focus of the preset project. Theorem 4. (Theorem XVII.5.3 from []). Let Y be a radom variable with characteristic fuctio ϕ Y (u) = e ψ Y (u), where u α Γ(3 α) C α( α) cos πα ( isg(u)(p q) ta πα ), α ψ Y (u) = u C π + isg(u)(p q) π ) l u, α = (53)

23 for some C > 0, p, q 0, p + q =, i.e. Y S α (γ, β, 0), where Γ(3 α) C γ α = α( α) cos πα, α C π ad β = p q (54), α = Let (X i ) i be i.i.d. radom variables whose distributio F satisfies either oe of the followig coditios: (a) U(x) = E(X { X x} ) is slowly varyig ad X is o-degeerate (b) U is regularly varyig of idex α for some 0 < α <, ad the tails of X are balaced, i.e. (30) holds. Let ( ) be a sequece of costats satisfyig (4). (i) If 0 < α <, the S d Y. (ii) If < α ad we assume that E(X ) = whe α =, the S E(X) d Y (iii) If α =, the S b [ ( )] d X Y, where b = E si. Corollary 4.3. (0 < α < ) Assume that P ( X > x) = x α L(x) where L is slowly varyig ad the tails are balaced, i.e. (30) holds. Choose ( ) such that (46) holds for some C > 0. The U(x) = x α l(x), where l = α L ad the α α coclusio of Theorem 4. holds with ψ Y (u) give by (53) with C = C α. Proof. We apply Theorem 4.. The fact that ( ) satisfies (4) follows by Lemma 4.9. Proof of Theorem 4.. Defie M such that M([ y, x]) = Cpx α + Cqy α, i.e. M = Cδ 0 if α = or M(dx) = ( α)cpx α+ (0, ) (x)dx + ( α)cq( x) α+ (,0) (x)dx if α <. Assume first that α <. The M = M α where M α is give by (4) with c + = ( α)cp ad c = ( α)cq. Note that c + + c α = C ad c + c c + + c = p q (55) Let W be a ifiitely divisible radom variable with ϕ W (u) = e ψ W (u) where e iux iu si x ψ W (u) = x M(dx) 3

24 Accordig to Lemma.8, W S α (γ, β, δ) where (γ, β, δ) are give by (5), (6), (8). Due to (55), formulas (5) ad (6) are exactly the same as (54). Hece, W δ S α (γ, β, 0), i.e. Y d = W δ. Therefore W d = Y + δ ad ψ W (u) = iuδ + ψ Y (u) Let F be the law of X ad M = x F (dx). (i) Assume that α <. We claim that i this case, e iux e iux x M (dx) x M(dx) (56) To see this, recall that M M properly (by Propositio 4.). Sice f(x) = e iux x is cotiuous o {x; x > δ}, e iux e iux lim x >δ x M (dx) = x >δ x M(dx) for ay δ > 0. (57) This still eeds to be show i more details. We see that e iux e iux lim δ 0 x M(dx) = x M(dx). x >δ Hece, to prove (56), it suffices to show that lim lim e iux e iux δ 0 x M (dx) x >δ x M (dx) = 0. (58) Note that, e iux x M (dx) x δ = x δ x δ e iux x x F (dx) = e iux F (dx) u = u E [ ] X { X aδ}. x δ (e iux )F (dx) x δ x F (dx) where we used the iequality e ia = (cos a ) + si a a for ay a. Usig Lemma 4.5.(ii) with β = (sice α < ), we obtai that E [ ] α X { X aδ} α (δ)p ( X > δ). 4

25 This gives us: x δ e iux x M (dx) α u α δp ( X > δ). Takig the upper limit whe teds to ifiity, we get: e iux lim x M (dx) α u δ lim α P ( X > δ) x δ Usig (49) ad (50) we obtai lim P ( X > δ) = C α α δ α. This gives us e iux lim x M (dx) u α α Cδ α. x δ Usig the fact that α > 0, we get that whe δ 0, the term o the right had side of the previous iequality coverges to zero. This cocludes the proof of (58) ad the proof of (56). Note that e iux x M (dx) = ad e iux x M(dx) = = E e iux x [ exp x F (dx) = ] (iu Xa ) e iux iu si x x = (e iux )F (dx) ( ) ] [ϕ X ua M(dx) + iu si x x M(dx) = ψ W (u) iuδ = ψ Y (u). ( ) ] Usig (56), we obtai: [ϕ X ψ Y (u), which is equivalet to the ua fact that S i the case α <. d Y (by emark 3.). This cocludes the proof of the theorem (ii) Assume that < α. We claim that, i this case e iux iux e iux iux x M (dx) x M(dx). (59) To see this, we use agai the fact that M M properly, so e iux iux e iux iux lim x t x M (dx) = x t x M(dx) (60) 5

26 for ay t > 0. We still eed to show this i more details. We see that e iux iux e iux iux lim t x M(dx) = x M(dx). x t Hece to show (59), it suffices to prove that: lim lim e iux iux e iux iux t x M (dx) x t x M (dx) = 0. Note that, x >t e iux iux x where we used the iequality M (dx) = = x >t x >t x >t e iux iux x x F (dx) (e iux iux)f (dx) ( + ) u e iux iux F (dx) x >t x F (dx) = ( + ) u E[ X { X >at}] e ia ia e ia + a ( + ) a. Usig Lemma 4.5.(i) with β = (sice α > ), we obtai that This gives us: x >t E[ X { X >at}] α α t U(t). e iux iux x M (dx) ( + ) α α u t Takig the upper limit whe teds to ifiity, we get: e iux iux lim x M (dx) ( + ) α α u t lim x >t U( t) a. U( t) a. U( t) Usig (4) we obtai lim a = Ct α. This gives us e iux iux lim x M (dx) ( + ) α α u Ct α. x >t 6

27 Usig the fact that α < 0, we get that whe t, the right had side of the previous iequality coverges to zero. This fiishes the proof of (59). Assume first that E(X) = 0. The e iux iux e iux iux x M (dx) = x x F (dx) [ ) ] = E exp (iu Xa iu Xa ( ) = [ϕ X i ua u ] E(X) ad e iux iux e iux iu si x x M(dx) = x = ψ W (u) iuδ = ψ Y (u). ( ) ] Usig (59), we obtai [ϕ X ψ W (u), i.e. ua M(dx) + iu S 3.). This fiishes the proof i the case α (, ] ad E(X) = 0. Assume ext that E(X) is arbitrary. We prove that ( ) ] = [ϕ X ua si x x x M(dx) d Y (by emark S E(X) d Y. (6) Writig E(X) = E(X { X >a}) + E(X { X a}), ad lettig b = E(X { X a}), we have We claim that S E(X) = S E(X { X a}) E(X { X >a}) = S b E(X { X >a}). (6) E [ ] X { X >a} x x M(dx) := δ. (63) To see this we use (57) with δ =. We get: e iux x M (dx) x x x e iux x M(dx). By takig the differece betwee (59) ad (60) (with t = ), we obtai: e iux iux e iux iux x M (dx) x M(dx). 7 x

28 Takig the differece betwee these two equatios, we obtai: iu x M (dx) = iu xf (dx) iu x M(dx). x x Due to (6) ad (63), (6) is equivalet to x S b d Y := Y + δ. (64) From (59), we kow that ( ) [ϕ X iu ua E(X) ] = = e iux iux x M (dx) e iux iux x M(dx) e iux iu si x x M(dx) si x x +iu M(dx) x = ψ W (u) iuδ = ψ Y (u), (65) usig defiitio (8) of δ ad the fact that Y = d W δ. Note that (64) is equivalet to ( ) } ϕ S a b (u) = {ϕ X e ua iub e ψ Y (u) = ϕ Y (u) To prove this, it suffices to show that { ( )} ( ( ) ]) u ϕ X exp [ϕ X ua sice the ( ) } ( ( ) ]) {ϕ X e ua iub exp [ϕ X iub e ua ψ Y (u) = e iuδ +ψ Y (u). For the last covergece, we used the fact that: ( ) ] ( ) [ϕ X iub = [ϕ X iu ua ua E(X ] { X }) ( ) = [ϕ X iu ua E(X) ] + iu E(X { X >}) = ψ Y (u) + iuδ, due to (65) ad (63). (66) 8

29 For the proof of (66), we refer to part (iii) below. i the case α (, ]. This cocludes the proof (iii) Assume that α =. Note that e iux iu si x x M (dx) = e iux iu si x x x F (dx) = (e iux iu si x)f (dx) ( ) )] = [ϕ X iue (si ua Xa ( ) ] = [ϕ X iub ua ad, sice i this case δ = 0, we have: e iux iu si x x M(dx) = ψ Y (u) = ψ W (u). By Propositio 4., it follows that ( ) ] [ϕ X iub ψ W (u). (67) ua To show that S b ϕ S a b = d W, we have to show that: {ϕ X ( ua ) e iub } e ψ W (u) = ϕ W (u). (68) To show (68), it suffices to prove (66), sice the ( ) } ( ( ) ]) {ϕ X e ua iub exp [ϕ X iub ua ad by (67), this last term coverges to e ψ W (u). The remaiig part of the proof is dedicated to (66). Note that relatio (66) ca be writte as ( + z ) e z, where z = ϕ X ( u ) 0, z C, z. We claim that z 0 implies ( + z ) e z. This implicatio still eeds to be prove. Hece it suffices to prove that ( ) u ϕ X 9 0 (69)

30 To prove (69), we use the fact that E X β < for β < (by Lemma 4.5.(i), sice α ). First ote that ( ) [ )] ( u ϕ X = E exp (iu ua exp iu x ) F (dx). Now we use the fact that e it t β, for ay t, β (0, ]. We get: ( ) u ϕ X β ux F (dx) u β This gives us ( ) u ϕ X a β 4 u β (E X β ) x β F (dx) = u β a β E X β. a β We claim that for ay δ > 0, there exists N δ such that C a α δ ( ) = O a β. (70) C a α+δ for ay N δ (7) To see this, recall that U(x) = x α l(x), where l is slowly varyig. By emark A.6, this implies that for ay δ > 0, there exists x δ such that x δ l(x) x δ for ay x x δ. The fact that implies that there exists N δ such that x δ for ay N δ. We the get: a δ U( ) a α+δ for ay N δ. a α δ By (4), C U() a U( ) a C U() a C U() a l( ) a δ for ay N δ. The C. This implies that there exist C, C > 0 such that C, for ay > 0. Therefore, for ay N δ a α δ a a α+δ a cocludes the proof of (7). From (7), we have that: a α+δ Hece, a β C β/(α+δ) ad a β = a α δ, which implies that C a α+δ = a α+δ, which implies that C a α δ ad. This ( ) /(α+δ), which gives us /(α+δ). C = C C β/(α+δ) β/(α+δ) 0 as (7) if < β α + δ i.e α + δ < β. Sice we also eed β <, we must choose δ such that α + δ <, i.e. 0 < δ < α. elatio (69) follows from (70) ad (7). This cocludes the proof i this case α =. C 30

31 I the case α (0, ), there is a alterative formulatio of Theorem 4., based o the represetatio of the characteristic fuctio of a stable radom variable give by Lemma.0. Theorem 4.4. Let (X i ) i be i.i.d. radom variables such that P ( X > x) = x α L(x) where L is slowly varyig ad 0 < α <, ad the tails are balaced i.e. (30) is satisfied for some p, q 0, p+q =. Let S = X i. Choose ( ) such that (46) holds for some costat C > 0. Let Y be a ifiitely divisible radom variable with characteristic fuctio { } ϕ Y (u) = exp (e iux iux { x } )ν α (dx) where ν α (dx) = C α[px α (0, ) (x) + q( x) α (,0) (x)]dx. The S d b Y, where b = E(X { X }) α Proof. By Lemma 4.9, ( ) satisfies (4) with C = C α. Hece C = α α C. Note that the measure ν α has the form (9) with i= c + = C pα = ( α)cp ad c = C qα = ( α)cq. By Lemma.0, Y S α (γ, β, δ ) where (γ, β) are give by (5), (6) ad δ is give by (0). Sice c + + c α = C ad c + c = p q, formulas (5) ad (6) c + + c are exactly the same as (54). X Let F be the law of ad M (dx) = x F (dx). We will use some of the results show i the proof of Theorem 4. with M α (dx) = x ν α (dx). We cosider three cases. ) Suppose that α <. By Theorem 4..(i), S d Y Sα (γ, β, 0). We claim that i this case, E(X { X a}) To see this, by (60), with t =, we get (e iux iux)f (dx) x 3 x x x ν α (dx) = δ. (73) (e iux iux) ν α (dx). (74)

32 Takig the differece betwee (56) ad (57) with δ =, we obtai (e iux )F (dx) (e iux ) ν α (dx). (75) x x Takig the differece betwee (74) ad (75), we get [ ] X E { X a} = xf (dx) This proves (73). Hece x x S E(X { X a}) δ + Y = Y. x ν α (dx) = δ. ) Suppose that α >. By relatio (64) i the proof of Theorem 4. S E(X { X a}) δ + Y = Y. 3) Suppose that α =. By Theorem 4..(iii), S ] E [si Xa d Y. We claim that ] E [si Xa E(X { X }) δ = (si x x { x } )ν α (dx). (76) To see this, ote that by (57) with δ =, (e iux )F (dx) x > (e iux )ν α (dx). By (60) with t =, (e iux iux)f (dx) x (e iux iux)ν α (dx). Takig the sum, we obtai: (e iux iux { x } )F (dx) (e iux iux { x } )ν α (dx). By (47), (e iux si x)f (dx) (e iux si x)ν α (dx). Takig the differece betwee the two previous equatios, we obtai: (si x x { x } )F (dx) (si x x { x } )ν α (dx). This proves (76). This gives us S E(X { X a}) δ + Y = Y. 3

33 emark 4.5. Uder the coditios of Theorem 4.4, we have: ] E [si Xa si x x M(dx) = si x ν α (dx) if α (0, ). (77) To see this, ote that ad ψ Y (u) = ψ (u) = = \{0} e iux iu si x x M (dx) e iux x M (dx) iu e iux iu si x x M(dx) = Usig (47) ad (56), we obtai: ] E [si Xa = si x x x F (dx) si x x x F (dx) e iux x M(dx) iu si x x M(dx). si x x M(dx) emark 4.6. Uder the coditios of Theorem 4.4 [ X E si X ] x si x x M(dx) = (x si x) ν α (dx) if α (, ]. To see this, ote that ad ψ (u) = = ψ Y (u) = = Usig (47) ad (59), we obtai: [ ( )] X X E si = e iux iu si x x M (dx) e iux iux x M (dx) iu e iux iu si x x M(dx) e iux iux x M(dx) iu \{0} x si x x x F (dx) x si x x x F (dx) x si x x M(dx) x si x x M(dx). 33

34 5 Simulatios o I this sectio we will use Theorem 4. to verify if certai distributios are i the domai of attractio of a stable distributio. We cosider a radom variable X with Pareto distributio of parameter α with distributio fuctio F. Let (X i ) i be i.i.d. copies of X. Let S = X i. By Corollary 4.3, for 0 < α <, X is i the domai of attractio of a stable distributio, i.e. S b d Y Sα (γ, β, 0), (where the parameters are give i Theorem 4.) if P ( X > x) = x α L(x) where L is a slowly varyig fuctio (78) ad the tails are balaced, i.e. (30) holds. We choose satisifyig (46). By Theorem 4., for α =, X is i the domai of attractio of a stable distributio if U(x) = E(X { X x} ) is slowly varyig ad X is o-degeerate. We choose satisfyig (4). Because the Pareto distributio is always positive, we get that (78) is equivalet to P (X > x) = x α L(x), which is i tur equivalet to F (x) = P (X > x) = x α L(x). Aother cosequece of the positiveess of X is that P (X > x) lim x P ( X > x) = ad lim P (X < x) = 0. So (30) holds with p = x P ( X > x) ad q = 0. Therefore β = p + q =. I this example, for 0 < α < we will = c, where c. The, choose{ L(x) F (x) = cx α. We the obtai: f(x) = cαx α, x > c 0 0, x c 0 for some c 0. { Takig c 0 =, we obtai X P areto(α), i.e. f(x) = αx α, x > 0, x. For α =, we have U(x) = E(X { X x} ) = l(x) for some slowly varyig fuctio l. We have: U(x) = x x y αy α dy = x i= y dy = l x = l(x). (79) We will ow fid the for 0 < α <, i.e. satisfyig (46). Takig C =, ad usig the positiveess of X, we get: P (X > ). We will choose such that P (X > ) =. Usig the fact that, for X P areto(α), P (X > x) = x α, we the obtai P (X > ) = a α =. Hece, = /α. α By Corollary 4.3, C = C α = α α. 34

35 We will ow fid the for α =. Usig (79) ad (4), we get: a = C l(). Takig C = we obtai the followig recurrece relatio: a = l( ) (80) ( Note that = exp ) W ( /) where W is the secod brach of the Lambert W fuctio. To compute the value of W we used the gsl package. With = 0000, we obtai: I the ext examples we will use the parameter α with values 0.5,,.5 ad. I each case, usig, we will simulate k = 0000 Pareto distributio samples of populatio = 0000 usig the VGAM package. We will the calcutate the sum = S(i) of values iside the sample to obtai S (i) for i k. Let W (i) b. Let W be the vector with W (i) as coordiates, the by Theorem 4., whe d k is very large, W Sα (γ, β, 0). To illustrate this result, we compute the desity fuctio of W by the histogram method o, ad the empirical distributio fuctio of W. We the superpose the graphs obtaied with the desity fuctio ad the distributio fuctio of a stable distributio with parameters (γ, β, 0). The graphs should coverge. Case : α = 0.5. ( Γ(3 α) I this case, Theorem 4. gives us b = 0 ad γ = C α( α) cos πα ) /α. We kow α = 0.5, the C = /3, hece γ We the compute W ad draw the two curves. To draw the stable curves, we use a Lévy distributio with parameters µ = 0 ad c = (because as see i [3], Lévy(µ, c) Stable 0.5 (, c, µ)). Therefore, by Theorem 4., we obtai: W = S d Y S 0.5 ( ,, 0) Lévy(0, ). We obtai the graphs (Y : red curve, W :black curve) i Figure Case : α =. I this case, Theorem 4. gives us E ( ( )) X si ad γ = C π. We kow 35

36 Figure : Illustratio of Stable Limit Theorem for α = 0.5 α =, the C =, hece γ = π. We ow calculate b : ( ( )) X ( x ) b = E si = si αx α dx /α /α ( x = si x ) dx = si y(y) (dy) = / si y dy y usig y = x ad computig the last itegral with = 0000 o Therefore, by Theorem 4., we obtai: W = S d Y Stable ( π,, 0) To draw the stable curve, we use the package stabledist. The fuctio uses the approach of J.P. Nola for geeral stable distributios. We obtai the graphs (Y : red curve, W :black curve) i Figure / Case 3: α =.5. I this case, Theorem 4. gives us b = E(X) ( Γ(3 α) ad γ = C α( α) cos πα ) /α. We kow α =.5, the C = 3, hece γ We ow calculate b : b = E(X) = /α =.5 /3 x.5 dx = 3 /3 36 xαx α dx

37 Figure : Illustratio of Stable Limit Theorem for α = Therefore, by Theorem 4., we obtai: W = /3 S 3 /3 d Y Stable.5 (.8457,, 0) The stable curve is draw usig the package stabledist. We obtai the graphs (Y : red curve, W :black curve) i Figure 3 Case 4: α =. I this case, Theorem 4. gives us b = E(X) ( Γ(3 α) ad γ = C α( α) cos πα ) /α. We kow α = ad we took C =, hece γ =. We ow calculate b : b = E(X) = xαx α dx = x dx = by Theorem 4., we obtai: W = S d Y Stable (,, 0) The stable curve is draw usig the package stabledist. We obtai the graphs (Y : red curve, W :black curve) i Figure 4 37

38 Figure 3: Illustratio of Stable Limit Theorem for α =.5 Figure 4: Illustratio of Stable Limit Theorem for α = 38

39 A Auxiliary esults Defiitio A.. A measure M is called caoical if it attributes fiite masses to fiite itervals ad the itegrals M + (x) = x y M(dy) ad M ( x) = are fiite for some (ad therefore all) x > 0. x M(dy) (8) y Defiitio A.. Let (M ) ad M be caoical measures. We say that (M ) coverges properly to M (ad we write M M properly) if: (i) M ([a, b]) M([a, b]) for ay a, b cotiuity poits of M, (ii)m + (x) M + (x) ad M ( x) M ( x) for ay x > 0 which is a cotiuity poit of M. Theorem A.3 (Theorem XVII.. of []). Let M be a caoical measure o, b ad ψ (u) = iub + + e iux iu si x x M (dx) u. (8) There exists a cotiuous fuctio ψ : C such that ψ (u) ψ(u) u if ad oly if there exist a caoical measure M o ad b such that I this case, ψ(u) = iub + Corollary A.4. Let M M properly ad b b. + e iux iu si x x M(dx) u. (83) ψ (u) = C (ϕ (u) iu b ) (84) where ϕ is the characteristic fuctio of a probability distributio F ad b, C. The there exists a cotiuous fuctio ψ : C such that ψ (u) ψ(u) for ay u if ad oly if C x F (dx) M(dx) ad C (β b ) b for some caoical measure M ad b where β = si xf (dx). I this case, ψ(u) is give by (83). Proof. Note that ψ (u) ca be writte i the form (8) with ad. The result follows by Theorem A.3. M (dx) = C x F (dx) b = C (β b ) 39

40 Defiitio A.5. (i) A fuctio L is slowly varyig (SV) if L(λx) lim = for ay λ > 0. x L(x) (ii) A fuctio U is regularly varyig with idex ρ (V ρ ) if for a slowly varyig fuctio L. U(x) = x ρ L(x) emark A.6. Let L be a slowly varyig fuctio. The for ay δ > 0 there exists x δ > 0 such that x δ < L(x) < x δ for ay x > x δ. Lemma A.7 (Lemma VIII.8.3 of []). Suppose that λ + λ ad. If U is a mootoe fuctio such that: lim λ U( x) = χ(x) exists for ay x D (D is a dese set) χ(x) is fiite ad positive for ay x I (I is a iterval), the U is regularly varyig of idex ρ ad χ(x) = Cx ρ, for some C > 0. Theorem A.8 (Theorem VIII.9..(i) of []). Let a > 0 ad < b < a. Defie the trucated momet fuctios U a ad V b by U a (x) = E( X a { X x} ) ad V b (x) = E( X b { X >x} ) Suppose that U a ( ) =. (i) If either U a or V b varies regularly the u a b V b (u) lim = c (85) u U a (u) for some 0 c. We write this limit uiquely i the form c = a α α b (86) for b α a with α = b if c =. (ii) If (85) holds for some c (0, ) the there exists a slowly varyig fuctio L such that U a (x) (α b)x a α L(x) ad V b (x) (a α)x b α L(x) where α [b, a] is chose such that (86) holds. 40

41 efereces [] Feller, William. A Itroductio to Probability Theory ad Its Applicatios. Joh Wiley, 97. [] Billigsley, Patrick. Covergece of Probability Measures. Joh Wiley, 999. [3] P. Nola, Joh. Stable Distributios: Models for Heavy Tailed Data [4] Gaudreau Lamarre, Pierre-Yves. oots of x = l(x). 0 4

Introduction to Extreme Value Theory Laurens de Haan, ISM Japan, Erasmus University Rotterdam, NL University of Lisbon, PT

Introduction to Extreme Value Theory Laurens de Haan, ISM Japan, Erasmus University Rotterdam, NL University of Lisbon, PT Itroductio to Extreme Value Theory Laures de Haa, ISM Japa, 202 Itroductio to Extreme Value Theory Laures de Haa Erasmus Uiversity Rotterdam, NL Uiversity of Lisbo, PT Itroductio to Extreme Value Theory

More information

Convergence of random variables. (telegram style notes) P.J.C. Spreij

Convergence of random variables. (telegram style notes) P.J.C. Spreij Covergece of radom variables (telegram style otes).j.c. Spreij this versio: September 6, 2005 Itroductio As we kow, radom variables are by defiitio measurable fuctios o some uderlyig measurable space

More information

7.1 Convergence of sequences of random variables

7.1 Convergence of sequences of random variables Chapter 7 Limit Theorems Throughout this sectio we will assume a probability space (, F, P), i which is defied a ifiite sequece of radom variables (X ) ad a radom variable X. The fact that for every ifiite

More information

Chapter 6 Infinite Series

Chapter 6 Infinite Series Chapter 6 Ifiite Series I the previous chapter we cosidered itegrals which were improper i the sese that the iterval of itegratio was ubouded. I this chapter we are goig to discuss a topic which is somewhat

More information

6.3 Testing Series With Positive Terms

6.3 Testing Series With Positive Terms 6.3. TESTING SERIES WITH POSITIVE TERMS 307 6.3 Testig Series With Positive Terms 6.3. Review of what is kow up to ow I theory, testig a series a i for covergece amouts to fidig the i= sequece of partial

More information

(A sequence also can be thought of as the list of function values attained for a function f :ℵ X, where f (n) = x n for n 1.) x 1 x N +k x N +4 x 3

(A sequence also can be thought of as the list of function values attained for a function f :ℵ X, where f (n) = x n for n 1.) x 1 x N +k x N +4 x 3 MATH 337 Sequeces Dr. Neal, WKU Let X be a metric space with distace fuctio d. We shall defie the geeral cocept of sequece ad limit i a metric space, the apply the results i particular to some special

More information

Infinite Sequences and Series

Infinite Sequences and Series Chapter 6 Ifiite Sequeces ad Series 6.1 Ifiite Sequeces 6.1.1 Elemetary Cocepts Simply speakig, a sequece is a ordered list of umbers writte: {a 1, a 2, a 3,...a, a +1,...} where the elemets a i represet

More information

7.1 Convergence of sequences of random variables

7.1 Convergence of sequences of random variables Chapter 7 Limit theorems Throughout this sectio we will assume a probability space (Ω, F, P), i which is defied a ifiite sequece of radom variables (X ) ad a radom variable X. The fact that for every ifiite

More information

Sequences and Series of Functions

Sequences and Series of Functions Chapter 6 Sequeces ad Series of Fuctios 6.1. Covergece of a Sequece of Fuctios Poitwise Covergece. Defiitio 6.1. Let, for each N, fuctio f : A R be defied. If, for each x A, the sequece (f (x)) coverges

More information

Chapter 3. Strong convergence. 3.1 Definition of almost sure convergence

Chapter 3. Strong convergence. 3.1 Definition of almost sure convergence Chapter 3 Strog covergece As poited out i the Chapter 2, there are multiple ways to defie the otio of covergece of a sequece of radom variables. That chapter defied covergece i probability, covergece i

More information

1 = δ2 (0, ), Y Y n nδ. , T n = Y Y n n. ( U n,k + X ) ( f U n,k + Y ) n 2n f U n,k + θ Y ) 2 E X1 2 X1

1 = δ2 (0, ), Y Y n nδ. , T n = Y Y n n. ( U n,k + X ) ( f U n,k + Y ) n 2n f U n,k + θ Y ) 2 E X1 2 X1 8. The cetral limit theorems 8.1. The cetral limit theorem for i.i.d. sequeces. ecall that C ( is N -separatig. Theorem 8.1. Let X 1, X,... be i.i.d. radom variables with EX 1 = ad EX 1 = σ (,. Suppose

More information

MASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.436J/15.085J Fall 2008 Lecture 19 11/17/2008 LAWS OF LARGE NUMBERS II THE STRONG LAW OF LARGE NUMBERS

MASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.436J/15.085J Fall 2008 Lecture 19 11/17/2008 LAWS OF LARGE NUMBERS II THE STRONG LAW OF LARGE NUMBERS MASSACHUSTTS INSTITUT OF TCHNOLOGY 6.436J/5.085J Fall 2008 Lecture 9 /7/2008 LAWS OF LARG NUMBRS II Cotets. The strog law of large umbers 2. The Cheroff boud TH STRONG LAW OF LARG NUMBRS While the weak

More information

MAS111 Convergence and Continuity

MAS111 Convergence and Continuity MAS Covergece ad Cotiuity Key Objectives At the ed of the course, studets should kow the followig topics ad be able to apply the basic priciples ad theorems therei to solvig various problems cocerig covergece

More information

Measure and Measurable Functions

Measure and Measurable Functions 3 Measure ad Measurable Fuctios 3.1 Measure o a Arbitrary σ-algebra Recall from Chapter 2 that the set M of all Lebesgue measurable sets has the followig properties: R M, E M implies E c M, E M for N implies

More information

lim za n n = z lim a n n.

lim za n n = z lim a n n. Lecture 6 Sequeces ad Series Defiitio 1 By a sequece i a set A, we mea a mappig f : N A. It is customary to deote a sequece f by {s } where, s := f(). A sequece {z } of (complex) umbers is said to be coverget

More information

Econ 325/327 Notes on Sample Mean, Sample Proportion, Central Limit Theorem, Chi-square Distribution, Student s t distribution 1.

Econ 325/327 Notes on Sample Mean, Sample Proportion, Central Limit Theorem, Chi-square Distribution, Student s t distribution 1. Eco 325/327 Notes o Sample Mea, Sample Proportio, Cetral Limit Theorem, Chi-square Distributio, Studet s t distributio 1 Sample Mea By Hiro Kasahara We cosider a radom sample from a populatio. Defiitio

More information

Lesson 10: Limits and Continuity

Lesson 10: Limits and Continuity www.scimsacademy.com Lesso 10: Limits ad Cotiuity SCIMS Academy 1 Limit of a fuctio The cocept of limit of a fuctio is cetral to all other cocepts i calculus (like cotiuity, derivative, defiite itegrals

More information

Notes 19 : Martingale CLT

Notes 19 : Martingale CLT Notes 9 : Martigale CLT Math 733-734: Theory of Probability Lecturer: Sebastie Roch Refereces: [Bil95, Chapter 35], [Roc, Chapter 3]. Sice we have ot ecoutered weak covergece i some time, we first recall

More information

Product measures, Tonelli s and Fubini s theorems For use in MAT3400/4400, autumn 2014 Nadia S. Larsen. Version of 13 October 2014.

Product measures, Tonelli s and Fubini s theorems For use in MAT3400/4400, autumn 2014 Nadia S. Larsen. Version of 13 October 2014. Product measures, Toelli s ad Fubii s theorems For use i MAT3400/4400, autum 2014 Nadia S. Larse Versio of 13 October 2014. 1. Costructio of the product measure The purpose of these otes is to preset the

More information

f n (x) f m (x) < ɛ/3 for all x A. By continuity of f n and f m we can find δ > 0 such that d(x, x 0 ) < δ implies that

f n (x) f m (x) < ɛ/3 for all x A. By continuity of f n and f m we can find δ > 0 such that d(x, x 0 ) < δ implies that Lecture 15 We have see that a sequece of cotiuous fuctios which is uiformly coverget produces a limit fuctio which is also cotiuous. We shall stregthe this result ow. Theorem 1 Let f : X R or (C) be a

More information

If a subset E of R contains no open interval, is it of zero measure? For instance, is the set of irrationals in [0, 1] is of measure zero?

If a subset E of R contains no open interval, is it of zero measure? For instance, is the set of irrationals in [0, 1] is of measure zero? 2 Lebesgue Measure I Chapter 1 we defied the cocept of a set of measure zero, ad we have observed that every coutable set is of measure zero. Here are some atural questios: If a subset E of R cotais a

More information

MASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.265/15.070J Fall 2013 Lecture 21 11/27/2013

MASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.265/15.070J Fall 2013 Lecture 21 11/27/2013 MASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.265/15.070J Fall 2013 Lecture 21 11/27/2013 Fuctioal Law of Large Numbers. Costructio of the Wieer Measure Cotet. 1. Additioal techical results o weak covergece

More information

2.2. Central limit theorem.

2.2. Central limit theorem. 36.. Cetral limit theorem. The most ideal case of the CLT is that the radom variables are iid with fiite variace. Although it is a special case of the more geeral Lideberg-Feller CLT, it is most stadard

More information

REAL ANALYSIS II: PROBLEM SET 1 - SOLUTIONS

REAL ANALYSIS II: PROBLEM SET 1 - SOLUTIONS REAL ANALYSIS II: PROBLEM SET 1 - SOLUTIONS 18th Feb, 016 Defiitio (Lipschitz fuctio). A fuctio f : R R is said to be Lipschitz if there exists a positive real umber c such that for ay x, y i the domai

More information

M17 MAT25-21 HOMEWORK 5 SOLUTIONS

M17 MAT25-21 HOMEWORK 5 SOLUTIONS M17 MAT5-1 HOMEWORK 5 SOLUTIONS 1. To Had I Cauchy Codesatio Test. Exercise 1: Applicatio of the Cauchy Codesatio Test Use the Cauchy Codesatio Test to prove that 1 diverges. Solutio 1. Give the series

More information

Definition 4.2. (a) A sequence {x n } in a Banach space X is a basis for X if. unique scalars a n (x) such that x = n. a n (x) x n. (4.

Definition 4.2. (a) A sequence {x n } in a Banach space X is a basis for X if. unique scalars a n (x) such that x = n. a n (x) x n. (4. 4. BASES I BAACH SPACES 39 4. BASES I BAACH SPACES Sice a Baach space X is a vector space, it must possess a Hamel, or vector space, basis, i.e., a subset {x γ } γ Γ whose fiite liear spa is all of X ad

More information

4. Partial Sums and the Central Limit Theorem

4. Partial Sums and the Central Limit Theorem 1 of 10 7/16/2009 6:05 AM Virtual Laboratories > 6. Radom Samples > 1 2 3 4 5 6 7 4. Partial Sums ad the Cetral Limit Theorem The cetral limit theorem ad the law of large umbers are the two fudametal theorems

More information

Fall 2013 MTH431/531 Real analysis Section Notes

Fall 2013 MTH431/531 Real analysis Section Notes Fall 013 MTH431/531 Real aalysis Sectio 8.1-8. Notes Yi Su 013.11.1 1. Defiitio of uiform covergece. We look at a sequece of fuctios f (x) ad study the coverget property. Notice we have two parameters

More information

62. Power series Definition 16. (Power series) Given a sequence {c n }, the series. c n x n = c 0 + c 1 x + c 2 x 2 + c 3 x 3 +

62. Power series Definition 16. (Power series) Given a sequence {c n }, the series. c n x n = c 0 + c 1 x + c 2 x 2 + c 3 x 3 + 62. Power series Defiitio 16. (Power series) Give a sequece {c }, the series c x = c 0 + c 1 x + c 2 x 2 + c 3 x 3 + is called a power series i the variable x. The umbers c are called the coefficiets of

More information

Solutions to HW Assignment 1

Solutions to HW Assignment 1 Solutios to HW: 1 Course: Theory of Probability II Page: 1 of 6 Uiversity of Texas at Austi Solutios to HW Assigmet 1 Problem 1.1. Let Ω, F, {F } 0, P) be a filtered probability space ad T a stoppig time.

More information

Limit distributions for products of sums

Limit distributions for products of sums Statistics & Probability Letters 62 (23) 93 Limit distributios for products of sums Yogcheg Qi Departmet of Mathematics ad Statistics, Uiversity of Miesota-Duluth, Campus Ceter 4, 7 Uiversity Drive, Duluth,

More information

Let us give one more example of MLE. Example 3. The uniform distribution U[0, θ] on the interval [0, θ] has p.d.f.

Let us give one more example of MLE. Example 3. The uniform distribution U[0, θ] on the interval [0, θ] has p.d.f. Lecture 5 Let us give oe more example of MLE. Example 3. The uiform distributio U[0, ] o the iterval [0, ] has p.d.f. { 1 f(x =, 0 x, 0, otherwise The likelihood fuctio ϕ( = f(x i = 1 I(X 1,..., X [0,

More information

Sequences. A Sequence is a list of numbers written in order.

Sequences. A Sequence is a list of numbers written in order. Sequeces A Sequece is a list of umbers writte i order. {a, a 2, a 3,... } The sequece may be ifiite. The th term of the sequece is the th umber o the list. O the list above a = st term, a 2 = 2 d term,

More information

7 Sequences of real numbers

7 Sequences of real numbers 40 7 Sequeces of real umbers 7. Defiitios ad examples Defiitio 7... A sequece of real umbers is a real fuctio whose domai is the set N of atural umbers. Let s : N R be a sequece. The the values of s are

More information

MAT1026 Calculus II Basic Convergence Tests for Series

MAT1026 Calculus II Basic Convergence Tests for Series MAT026 Calculus II Basic Covergece Tests for Series Egi MERMUT 202.03.08 Dokuz Eylül Uiversity Faculty of Sciece Departmet of Mathematics İzmir/TURKEY Cotets Mootoe Covergece Theorem 2 2 Series of Real

More information

Seunghee Ye Ma 8: Week 5 Oct 28

Seunghee Ye Ma 8: Week 5 Oct 28 Week 5 Summary I Sectio, we go over the Mea Value Theorem ad its applicatios. I Sectio 2, we will recap what we have covered so far this term. Topics Page Mea Value Theorem. Applicatios of the Mea Value

More information

Distribution of Random Samples & Limit theorems

Distribution of Random Samples & Limit theorems STAT/MATH 395 A - PROBABILITY II UW Witer Quarter 2017 Néhémy Lim Distributio of Radom Samples & Limit theorems 1 Distributio of i.i.d. Samples Motivatig example. Assume that the goal of a study is to

More information

Sequences and Limits

Sequences and Limits Chapter Sequeces ad Limits Let { a } be a sequece of real or complex umbers A ecessary ad sufficiet coditio for the sequece to coverge is that for ay ɛ > 0 there exists a iteger N > 0 such that a p a q

More information

Integrable Functions. { f n } is called a determining sequence for f. If f is integrable with respect to, then f d does exist as a finite real number

Integrable Functions. { f n } is called a determining sequence for f. If f is integrable with respect to, then f d does exist as a finite real number MATH 532 Itegrable Fuctios Dr. Neal, WKU We ow shall defie what it meas for a measurable fuctio to be itegrable, show that all itegral properties of simple fuctios still hold, ad the give some coditios

More information

Sequences. Notation. Convergence of a Sequence

Sequences. Notation. Convergence of a Sequence Sequeces A sequece is essetially just a list. Defiitio (Sequece of Real Numbers). A sequece of real umbers is a fuctio Z (, ) R for some real umber. Do t let the descriptio of the domai cofuse you; it

More information

We are mainly going to be concerned with power series in x, such as. (x)} converges - that is, lims N n

We are mainly going to be concerned with power series in x, such as. (x)} converges - that is, lims N n Review of Power Series, Power Series Solutios A power series i x - a is a ifiite series of the form c (x a) =c +c (x a)+(x a) +... We also call this a power series cetered at a. Ex. (x+) is cetered at

More information

Chapter 6 Principles of Data Reduction

Chapter 6 Principles of Data Reduction Chapter 6 for BST 695: Special Topics i Statistical Theory. Kui Zhag, 0 Chapter 6 Priciples of Data Reductio Sectio 6. Itroductio Goal: To summarize or reduce the data X, X,, X to get iformatio about a

More information

Advanced Analysis. Min Yan Department of Mathematics Hong Kong University of Science and Technology

Advanced Analysis. Min Yan Department of Mathematics Hong Kong University of Science and Technology Advaced Aalysis Mi Ya Departmet of Mathematics Hog Kog Uiversity of Sciece ad Techology September 3, 009 Cotets Limit ad Cotiuity 7 Limit of Sequece 8 Defiitio 8 Property 3 3 Ifiity ad Ifiitesimal 8 4

More information

s = and t = with C ij = A i B j F. (i) Note that cs = M and so ca i µ(a i ) I E (cs) = = c a i µ(a i ) = ci E (s). (ii) Note that s + t = M and so

s = and t = with C ij = A i B j F. (i) Note that cs = M and so ca i µ(a i ) I E (cs) = = c a i µ(a i ) = ci E (s). (ii) Note that s + t = M and so 3 From the otes we see that the parts of Theorem 4. that cocer us are: Let s ad t be two simple o-egative F-measurable fuctios o X, F, µ ad E, F F. The i I E cs ci E s for all c R, ii I E s + t I E s +

More information

An Introduction to Randomized Algorithms

An Introduction to Randomized Algorithms A Itroductio to Radomized Algorithms The focus of this lecture is to study a radomized algorithm for quick sort, aalyze it usig probabilistic recurrece relatios, ad also provide more geeral tools for aalysis

More information

1 Convergence in Probability and the Weak Law of Large Numbers

1 Convergence in Probability and the Weak Law of Large Numbers 36-752 Advaced Probability Overview Sprig 2018 8. Covergece Cocepts: i Probability, i L p ad Almost Surely Istructor: Alessadro Rialdo Associated readig: Sec 2.4, 2.5, ad 4.11 of Ash ad Doléas-Dade; Sec

More information

1+x 1 + α+x. x = 2(α x2 ) 1+x

1+x 1 + α+x. x = 2(α x2 ) 1+x Math 2030 Homework 6 Solutios # [Problem 5] For coveiece we let α lim sup a ad β lim sup b. Without loss of geerality let us assume that α β. If α the by assumptio β < so i this case α + β. By Theorem

More information

Beurling Integers: Part 2

Beurling Integers: Part 2 Beurlig Itegers: Part 2 Isomorphisms Devi Platt July 11, 2015 1 Prime Factorizatio Sequeces I the last article we itroduced the Beurlig geeralized itegers, which ca be represeted as a sequece of real umbers

More information

SOME THEORY AND PRACTICE OF STATISTICS by Howard G. Tucker

SOME THEORY AND PRACTICE OF STATISTICS by Howard G. Tucker SOME THEORY AND PRACTICE OF STATISTICS by Howard G. Tucker CHAPTER 9. POINT ESTIMATION 9. Covergece i Probability. The bases of poit estimatio have already bee laid out i previous chapters. I chapter 5

More information

MATH 112: HOMEWORK 6 SOLUTIONS. Problem 1: Rudin, Chapter 3, Problem s k < s k < 2 + s k+1

MATH 112: HOMEWORK 6 SOLUTIONS. Problem 1: Rudin, Chapter 3, Problem s k < s k < 2 + s k+1 MATH 2: HOMEWORK 6 SOLUTIONS CA PRO JIRADILOK Problem. If s = 2, ad Problem : Rudi, Chapter 3, Problem 3. s + = 2 + s ( =, 2, 3,... ), prove that {s } coverges, ad that s < 2 for =, 2, 3,.... Proof. The

More information

1 Introduction to reducing variance in Monte Carlo simulations

1 Introduction to reducing variance in Monte Carlo simulations Copyright c 010 by Karl Sigma 1 Itroductio to reducig variace i Mote Carlo simulatios 11 Review of cofidece itervals for estimatig a mea I statistics, we estimate a ukow mea µ = E(X) of a distributio by

More information

MA131 - Analysis 1. Workbook 3 Sequences II

MA131 - Analysis 1. Workbook 3 Sequences II MA3 - Aalysis Workbook 3 Sequeces II Autum 2004 Cotets 2.8 Coverget Sequeces........................ 2.9 Algebra of Limits......................... 2 2.0 Further Useful Results........................

More information

Topic 9: Sampling Distributions of Estimators

Topic 9: Sampling Distributions of Estimators Topic 9: Samplig Distributios of Estimators Course 003, 2018 Page 0 Samplig distributios of estimators Sice our estimators are statistics (particular fuctios of radom variables), their distributio ca be

More information

Math 2784 (or 2794W) University of Connecticut

Math 2784 (or 2794W) University of Connecticut ORDERS OF GROWTH PAT SMITH Math 2784 (or 2794W) Uiversity of Coecticut Date: Mar. 2, 22. ORDERS OF GROWTH. Itroductio Gaiig a ituitive feel for the relative growth of fuctios is importat if you really

More information

The natural exponential function

The natural exponential function The atural expoetial fuctio Attila Máté Brookly College of the City Uiversity of New York December, 205 Cotets The atural expoetial fuctio for real x. Beroulli s iequality.....................................2

More information

LECTURE 8: ASYMPTOTICS I

LECTURE 8: ASYMPTOTICS I LECTURE 8: ASYMPTOTICS I We are iterested i the properties of estimators as. Cosider a sequece of radom variables {, X 1}. N. M. Kiefer, Corell Uiversity, Ecoomics 60 1 Defiitio: (Weak covergece) A sequece

More information

1. By using truth tables prove that, for all statements P and Q, the statement

1. By using truth tables prove that, for all statements P and Q, the statement Author: Satiago Salazar Problems I: Mathematical Statemets ad Proofs. By usig truth tables prove that, for all statemets P ad Q, the statemet P Q ad its cotrapositive ot Q (ot P) are equivalet. I example.2.3

More information

Complex Analysis Spring 2001 Homework I Solution

Complex Analysis Spring 2001 Homework I Solution Complex Aalysis Sprig 2001 Homework I Solutio 1. Coway, Chapter 1, sectio 3, problem 3. Describe the set of poits satisfyig the equatio z a z + a = 2c, where c > 0 ad a R. To begi, we see from the triagle

More information

Sequences A sequence of numbers is a function whose domain is the positive integers. We can see that the sequence

Sequences A sequence of numbers is a function whose domain is the positive integers. We can see that the sequence Sequeces A sequece of umbers is a fuctio whose domai is the positive itegers. We ca see that the sequece 1, 1, 2, 2, 3, 3,... is a fuctio from the positive itegers whe we write the first sequece elemet

More information

Probability 2 - Notes 10. Lemma. If X is a random variable and g(x) 0 for all x in the support of f X, then P(g(X) 1) E[g(X)].

Probability 2 - Notes 10. Lemma. If X is a random variable and g(x) 0 for all x in the support of f X, then P(g(X) 1) E[g(X)]. Probability 2 - Notes 0 Some Useful Iequalities. Lemma. If X is a radom variable ad g(x 0 for all x i the support of f X, the P(g(X E[g(X]. Proof. (cotiuous case P(g(X Corollaries x:g(x f X (xdx x:g(x

More information

MATH4822E FOURIER ANALYSIS AND ITS APPLICATIONS

MATH4822E FOURIER ANALYSIS AND ITS APPLICATIONS MATH48E FOURIER ANALYSIS AND ITS APPLICATIONS 7.. Cesàro summability. 7. Summability methods Arithmetic meas. The followig idea is due to the Italia geometer Eresto Cesàro (859-96). He shows that eve if

More information

EE 4TM4: Digital Communications II Probability Theory

EE 4TM4: Digital Communications II Probability Theory 1 EE 4TM4: Digital Commuicatios II Probability Theory I. RANDOM VARIABLES A radom variable is a real-valued fuctio defied o the sample space. Example: Suppose that our experimet cosists of tossig two fair

More information

1 Approximating Integrals using Taylor Polynomials

1 Approximating Integrals using Taylor Polynomials Seughee Ye Ma 8: Week 7 Nov Week 7 Summary This week, we will lear how we ca approximate itegrals usig Taylor series ad umerical methods. Topics Page Approximatig Itegrals usig Taylor Polyomials. Defiitios................................................

More information

Comparison Study of Series Approximation. and Convergence between Chebyshev. and Legendre Series

Comparison Study of Series Approximation. and Convergence between Chebyshev. and Legendre Series Applied Mathematical Scieces, Vol. 7, 03, o. 6, 3-337 HIKARI Ltd, www.m-hikari.com http://d.doi.org/0.988/ams.03.3430 Compariso Study of Series Approimatio ad Covergece betwee Chebyshev ad Legedre Series

More information

Lecture 7: Properties of Random Samples

Lecture 7: Properties of Random Samples Lecture 7: Properties of Radom Samples 1 Cotiued From Last Class Theorem 1.1. Let X 1, X,...X be a radom sample from a populatio with mea µ ad variace σ

More information

A sequence of numbers is a function whose domain is the positive integers. We can see that the sequence

A sequence of numbers is a function whose domain is the positive integers. We can see that the sequence Sequeces A sequece of umbers is a fuctio whose domai is the positive itegers. We ca see that the sequece,, 2, 2, 3, 3,... is a fuctio from the positive itegers whe we write the first sequece elemet as

More information

Mathematics 170B Selected HW Solutions.

Mathematics 170B Selected HW Solutions. Mathematics 17B Selected HW Solutios. F 4. Suppose X is B(,p). (a)fidthemometgeeratigfuctiom (s)of(x p)/ p(1 p). Write q = 1 p. The MGF of X is (pe s + q), sice X ca be writte as the sum of idepedet Beroulli

More information

Sequences and Series

Sequences and Series Sequeces ad Series Sequeces of real umbers. Real umber system We are familiar with atural umbers ad to some extet the ratioal umbers. While fidig roots of algebraic equatios we see that ratioal umbers

More information

Topic 9: Sampling Distributions of Estimators

Topic 9: Sampling Distributions of Estimators Topic 9: Samplig Distributios of Estimators Course 003, 2018 Page 0 Samplig distributios of estimators Sice our estimators are statistics (particular fuctios of radom variables), their distributio ca be

More information

MASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.265/15.070J Fall 2013 Lecture 3 9/11/2013. Large deviations Theory. Cramér s Theorem

MASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.265/15.070J Fall 2013 Lecture 3 9/11/2013. Large deviations Theory. Cramér s Theorem MASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.265/5.070J Fall 203 Lecture 3 9//203 Large deviatios Theory. Cramér s Theorem Cotet.. Cramér s Theorem. 2. Rate fuctio ad properties. 3. Chage of measure techique.

More information

A Proof of Birkhoff s Ergodic Theorem

A Proof of Birkhoff s Ergodic Theorem A Proof of Birkhoff s Ergodic Theorem Joseph Hora September 2, 205 Itroductio I Fall 203, I was learig the basics of ergodic theory, ad I came across this theorem. Oe of my supervisors, Athoy Quas, showed

More information

Topic 9: Sampling Distributions of Estimators

Topic 9: Sampling Distributions of Estimators Topic 9: Samplig Distributios of Estimators Course 003, 2016 Page 0 Samplig distributios of estimators Sice our estimators are statistics (particular fuctios of radom variables), their distributio ca be

More information

Lecture 8: Convergence of transformations and law of large numbers

Lecture 8: Convergence of transformations and law of large numbers Lecture 8: Covergece of trasformatios ad law of large umbers Trasformatio ad covergece Trasformatio is a importat tool i statistics. If X coverges to X i some sese, we ofte eed to check whether g(x ) coverges

More information

1 Lecture 2: Sequence, Series and power series (8/14/2012)

1 Lecture 2: Sequence, Series and power series (8/14/2012) Summer Jump-Start Program for Aalysis, 202 Sog-Yig Li Lecture 2: Sequece, Series ad power series (8/4/202). More o sequeces Example.. Let {x } ad {y } be two bouded sequeces. Show lim sup (x + y ) lim

More information

It is always the case that unions, intersections, complements, and set differences are preserved by the inverse image of a function.

It is always the case that unions, intersections, complements, and set differences are preserved by the inverse image of a function. MATH 532 Measurable Fuctios Dr. Neal, WKU Throughout, let ( X, F, µ) be a measure space ad let (!, F, P ) deote the special case of a probability space. We shall ow begi to study real-valued fuctios defied

More information

Lecture 19: Convergence

Lecture 19: Convergence Lecture 19: Covergece Asymptotic approach I statistical aalysis or iferece, a key to the success of fidig a good procedure is beig able to fid some momets ad/or distributios of various statistics. I may

More information

Ma 4121: Introduction to Lebesgue Integration Solutions to Homework Assignment 5

Ma 4121: Introduction to Lebesgue Integration Solutions to Homework Assignment 5 Ma 42: Itroductio to Lebesgue Itegratio Solutios to Homework Assigmet 5 Prof. Wickerhauser Due Thursday, April th, 23 Please retur your solutios to the istructor by the ed of class o the due date. You

More information

Random Models. Tusheng Zhang. February 14, 2013

Random Models. Tusheng Zhang. February 14, 2013 Radom Models Tusheg Zhag February 14, 013 1 Radom Walks Let me describe the model. Radom walks are used to describe the motio of a movig particle (object). Suppose that a particle (object) moves alog the

More information

Math Solutions to homework 6

Math Solutions to homework 6 Math 175 - Solutios to homework 6 Cédric De Groote November 16, 2017 Problem 1 (8.11 i the book): Let K be a compact Hermitia operator o a Hilbert space H ad let the kerel of K be {0}. Show that there

More information

3. Z Transform. Recall that the Fourier transform (FT) of a DT signal xn [ ] is ( ) [ ] = In order for the FT to exist in the finite magnitude sense,

3. Z Transform. Recall that the Fourier transform (FT) of a DT signal xn [ ] is ( ) [ ] = In order for the FT to exist in the finite magnitude sense, 3. Z Trasform Referece: Etire Chapter 3 of text. Recall that the Fourier trasform (FT) of a DT sigal x [ ] is ω ( ) [ ] X e = j jω k = xe I order for the FT to exist i the fiite magitude sese, S = x [

More information

This section is optional.

This section is optional. 4 Momet Geeratig Fuctios* This sectio is optioal. The momet geeratig fuctio g : R R of a radom variable X is defied as g(t) = E[e tx ]. Propositio 1. We have g () (0) = E[X ] for = 1, 2,... Proof. Therefore

More information

Additional Notes on Power Series

Additional Notes on Power Series Additioal Notes o Power Series Mauela Girotti MATH 37-0 Advaced Calculus of oe variable Cotets Quick recall 2 Abel s Theorem 2 3 Differetiatio ad Itegratio of Power series 4 Quick recall We recall here

More information

CSE 1400 Applied Discrete Mathematics Number Theory and Proofs

CSE 1400 Applied Discrete Mathematics Number Theory and Proofs CSE 1400 Applied Discrete Mathematics Number Theory ad Proofs Departmet of Computer Scieces College of Egieerig Florida Tech Sprig 01 Problems for Number Theory Backgroud Number theory is the brach of

More information

Math 341 Lecture #31 6.5: Power Series

Math 341 Lecture #31 6.5: Power Series Math 341 Lecture #31 6.5: Power Series We ow tur our attetio to a particular kid of series of fuctios, amely, power series, f(x = a x = a 0 + a 1 x + a 2 x 2 + where a R for all N. I terms of a series

More information

MATH301 Real Analysis (2008 Fall) Tutorial Note #7. k=1 f k (x) converges pointwise to S(x) on E if and

MATH301 Real Analysis (2008 Fall) Tutorial Note #7. k=1 f k (x) converges pointwise to S(x) on E if and MATH01 Real Aalysis (2008 Fall) Tutorial Note #7 Sequece ad Series of fuctio 1: Poitwise Covergece ad Uiform Covergece Part I: Poitwise Covergece Defiitio of poitwise covergece: A sequece of fuctios f

More information

Ma 530 Introduction to Power Series

Ma 530 Introduction to Power Series Ma 530 Itroductio to Power Series Please ote that there is material o power series at Visual Calculus. Some of this material was used as part of the presetatio of the topics that follow. What is a Power

More information

Output Analysis and Run-Length Control

Output Analysis and Run-Length Control IEOR E4703: Mote Carlo Simulatio Columbia Uiversity c 2017 by Marti Haugh Output Aalysis ad Ru-Legth Cotrol I these otes we describe how the Cetral Limit Theorem ca be used to costruct approximate (1 α%

More information

2.1. Convergence in distribution and characteristic functions.

2.1. Convergence in distribution and characteristic functions. 3 Chapter 2. Cetral Limit Theorem. Cetral limit theorem, or DeMoivre-Laplace Theorem, which also implies the wea law of large umbers, is the most importat theorem i probability theory ad statistics. For

More information

Random Matrices with Blocks of Intermediate Scale Strongly Correlated Band Matrices

Random Matrices with Blocks of Intermediate Scale Strongly Correlated Band Matrices Radom Matrices with Blocks of Itermediate Scale Strogly Correlated Bad Matrices Jiayi Tog Advisor: Dr. Todd Kemp May 30, 07 Departmet of Mathematics Uiversity of Califoria, Sa Diego Cotets Itroductio Notatio

More information

Appendix: The Laplace Transform

Appendix: The Laplace Transform Appedix: The Laplace Trasform The Laplace trasform is a powerful method that ca be used to solve differetial equatio, ad other mathematical problems. Its stregth lies i the fact that it allows the trasformatio

More information

Chapter 10: Power Series

Chapter 10: Power Series Chapter : Power Series 57 Chapter Overview: Power Series The reaso series are part of a Calculus course is that there are fuctios which caot be itegrated. All power series, though, ca be itegrated because

More information

Math 220A Fall 2007 Homework #2. Will Garner A

Math 220A Fall 2007 Homework #2. Will Garner A Math 0A Fall 007 Homewor # Will Garer Pg 3 #: Show that {cis : a o-egative iteger} is dese i T = {z œ : z = }. For which values of q is {cis(q): a o-egative iteger} dese i T? To show that {cis : a o-egative

More information

CHAPTER 10 INFINITE SEQUENCES AND SERIES

CHAPTER 10 INFINITE SEQUENCES AND SERIES CHAPTER 10 INFINITE SEQUENCES AND SERIES 10.1 Sequeces 10.2 Ifiite Series 10.3 The Itegral Tests 10.4 Compariso Tests 10.5 The Ratio ad Root Tests 10.6 Alteratig Series: Absolute ad Coditioal Covergece

More information

MASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.265/15.070J Fall 2013 Lecture 2 9/9/2013. Large Deviations for i.i.d. Random Variables

MASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.265/15.070J Fall 2013 Lecture 2 9/9/2013. Large Deviations for i.i.d. Random Variables MASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.265/15.070J Fall 2013 Lecture 2 9/9/2013 Large Deviatios for i.i.d. Radom Variables Cotet. Cheroff boud usig expoetial momet geeratig fuctios. Properties of a momet

More information

Expectation and Variance of a random variable

Expectation and Variance of a random variable Chapter 11 Expectatio ad Variace of a radom variable The aim of this lecture is to defie ad itroduce mathematical Expectatio ad variace of a fuctio of discrete & cotiuous radom variables ad the distributio

More information

INFINITE SEQUENCES AND SERIES

INFINITE SEQUENCES AND SERIES INFINITE SEQUENCES AND SERIES INFINITE SEQUENCES AND SERIES I geeral, it is difficult to fid the exact sum of a series. We were able to accomplish this for geometric series ad the series /[(+)]. This is

More information

Math 155 (Lecture 3)

Math 155 (Lecture 3) Math 55 (Lecture 3) September 8, I this lecture, we ll cosider the aswer to oe of the most basic coutig problems i combiatorics Questio How may ways are there to choose a -elemet subset of the set {,,,

More information

Series III. Chapter Alternating Series

Series III. Chapter Alternating Series Chapter 9 Series III With the exceptio of the Null Sequece Test, all the tests for series covergece ad divergece that we have cosidered so far have dealt oly with series of oegative terms. Series with

More information

Math 113 Exam 3 Practice

Math 113 Exam 3 Practice Math Exam Practice Exam will cover.-.9. This sheet has three sectios. The first sectio will remid you about techiques ad formulas that you should kow. The secod gives a umber of practice questios for you

More information

1.3 Convergence Theorems of Fourier Series. k k k k. N N k 1. With this in mind, we state (without proof) the convergence of Fourier series.

1.3 Convergence Theorems of Fourier Series. k k k k. N N k 1. With this in mind, we state (without proof) the convergence of Fourier series. .3 Covergece Theorems of Fourier Series I this sectio, we preset the covergece of Fourier series. A ifiite sum is, by defiitio, a limit of partial sums, that is, a cos( kx) b si( kx) lim a cos( kx) b si(

More information