The number of minima in a discrete sample

Hele tekst

(1)The number of minima in a discrete sample Citation for published version (APA): Steutel, F. W. (1998). The number of minima in a discrete sample. (Memorandum COSOR; Vol. 9802). Technische Universiteit Eindhoven.. Document status and date: Published: 01/01/1998 Document Version: Publisher’s PDF, also known as Version of Record (includes final page, issue and volume numbers) Please check the document version of this publication: • A submitted manuscript is the version of the article upon submission and before peer-review. There can be important differences between the submitted version and the official published version of record. People interested in the research are advised to contact the author for the final version of the publication, or visit the DOI to the publisher's website. • The final author version and the galley proof are versions of the publication after peer review. • The final published version features the final layout of the paper including the volume, issue and page numbers. Link to publication. General rights Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights. • Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying the publication in the public portal. If the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license above, please follow below link for the End User Agreement: www.tue.nl/taverne. Take down policy If you believe that this document breaches copyright please contact us at: openaccess@tue.nl providing details and we will investigate your claim.. Download date: 10. Sep. 2021.

(2) Memorandum COSOR 98-02, 1998, Eindhoven University of Technology. The number of minima in a discrete sample F.W. Steutel. Abstract. The number of times is considered that the minimum occurs in a sample from a discrete distribution. The special case of the geometric distribution is considered in some detail, and applied to the computation of the expected maximum of a sample from the Cantor distribution.. 1. Introduction and summary Brands et al. (1994) and Baryshnikov et al. (1995) consider the number Kn of values equal to the maximum in a sample from a distribution on the nonnegative integers. For n ! 1, in some cases Kn tends to 1, in many cases it does not converge, and Kn tends to in nity, if the support of the distribution is nite. This last case is rather simple, and has not been given much attention. It is equivalent to the case where the number of minima of the sample is considered. In this note we look at this quantity, which we also denote by Kn. In Section 2 the general case is considered brie

(3) y. Section deals in some more detail with the case of the geometric distribution, and in Section 4 these results are applied to a problem, a special case of which appeared in the Problem Section of the American Mathematical Monthly (see Diamond and Reznick (1997) a somewhat similar problem is considered in van Harn (1988)).. 2. The general case Let M1 : : : Mn be independent and distributed as M with. P (M = j ) = pj P (M j ) = Pj . for j = 1 2 : : : . De ne and. Nn = min(M1 : : : Mn) Kn = #fj : Mj = Nng:. Department of Mathematics and Computing Science, Eindhoven University of Technology, Eindhoven, The Netherlands. 1.

(4) We are interested in the distribution of (Nn Kn). On account of symmetry one easily veri es the following results. Proposition 1. !. P (Nn = r Kn = k) = nk pkr (1 ; Pr )n;k P (Nn = r) = (1 ;! Pr;1)n ; (1 ; Pr )n 1 X n P (Kn = k) = k pkr (1 ; Pr )n;k 1. where P0 = 0. Corollary 1. !. P (Kn = k) = nk pk1 (1 ; p1)n;k (1 + e(n)) EKn = np1 (1 + e(n)). as n ! 1, where e(n) denotes an exponentially small contribution.. Of course, the result in Corollary 1 is not very surprising for large n the minimum will almost certainly be one, and Knwill be the number of ones in the sample, which is binomially distributed with succes probability p1.. 3. The geometric case. Now let M have a geometric distrtibution, i.e., let p 2 (0 1) and let. pj = pj;1(1 ; p) 1 ; Pj = pj j = 1 2 : : : : Whereas in general Nn and Kn are dependent, in this case they are not.. Proposition 2 If. independent, with. (1). M has a geometric distribution as given by (1), then Nn and Kn are P (Nn = r Kn = k) = (1 ;. !. p)rn n ( p )k : k 1;p. Proof. Follows directly from Proposition 1 with (1), and from the product form of the formula above.. 2.

(5) Corollary 2. !. k n;k P (Kn = k) = nk (1 ;1 ;p) ppn P (Nn = r) = pn(r;1) (1 ; pn). n n EzKn = (p + (11;;pp)zn) ; p . 4. An application. pn z: EzNn = 11;;zp n. In Problem 10621 (1997) one is requested to compute 1 ; mn , where Z1 mn = 1 ; F n(x)dx 0. (2) (3). (4). whith F a Cantor distribution function (with p = 1=2). This means that F is the distribution function of X given by 1 X X = ( 31 )j Yj (5) j =1 where Y1 Y2 : : : are independent and distributed as Y with with p 2 (0 1). Clearly,. P (Y = 0) = 1 ; P (Y = 2) = p. mn = Emax(X1 : : : Xn) with the X 's independent and distributed as X above. It follows from (5) that X has the form X = 0:00002 : : : where the number of zeroes preceding the rst 2 equals M ; 1, with M geometrically distributed as in (1) M indicates the position of the rst 2 in the triadic expansion of X . As a result we can apply the formulas derived in Section 3, and so we have max(X1 : : : Xn) =d 2( 31 )Nn + ( 13 )Nn max(X1 : : : XKn ) with Nn and Kn as in Section 3, and Nn Kn,and X1 X2 : : : independent. That is max(X1 : : : Xn) equals that Xj for which the rst 2 comes rst and for which the number represented by the remaining digits is maximal. Taking expectations we obtain mn = 2E( 31 )Nn + E( 13 )Nn EmKn 3.

(6) where we used the independence of Nn Kn and the X 's. By (3) of Corollary 2, mn can be expressed as n 1 1 ; pn Z 1 2 1 ; p mn = 3 1 ; 1 pn + 3 1 ; 1 pn E (1 ; F Kn (x))dx: Using (2) and (4) we get. 3. 3. 0. ; pn + 1 ; pn Z 1f1 ; (p + (1 ; p)F (x))n ; pn gdxg = mn = 2 31 ; pn 3 ; pn 1 ; pn 0. Z1 n 1 1 ; p = 2 3 ; pn + 3 ; pn f1 ; (p + (1 ; p)F (x)ngdx = 0 n n! n X 1 ; p 1 n;k (1 ; p)k mk : = 2 3 ; pn + 3 ; pn p k k=1. (6). From (6) we compute the rst few values of mn. For general p we obtain m1 = 1 ; p, m2 = (1+ p ; 3p2 + p3)=(1+ p ; p2), m3 = (2+5p ; 2p2 ; 17p3 +16p4 ; 4p5 )=(2+5p ; 2p2 ; 6p3 +3p4). For p = 1=2 we nd m1 = 1=2, m2 = 7=10, m3 = 4=5 and m4 = 197=230. The general expression for m4 involv polynomials of degrees nine and eight in the numerator and denominator, respectively for p = 1=2 the values of mn can easily be calculated we give m5 = 41=46, M6 = 799=874, and m7 = 8129=8740. Of course, mn tends to one as n ! 1.. References Baryshnikov, Y., Eisenberg, B. and Stengle, G., A necessary and sucient condition for the existence of the limiting probability of a tie for rst place, Statist. Prob. Letters 23 (1995), 203- 209. Brands, J.A.M., Steutel, F.W. and Wilms, R.J.G., On the number of maxima in a discrete sample, Statist. Prob. Letters 20 (1994), 209-217. Diamond, H.G. and Reznick, B. (1997), Problem 10621, American Mathematical Monthly 104, Nr. 9, p. 870. van Harn, K. (1988), Problem 208, solution by D. Gilat, Statistica Neerlandica 42, 70-71.. 4.

(7)

No results found