NONPARAMETRIC ESTIMATION FOR MIDDLE-CENSORED DATA
|
|
- Leslie Nicholson
- 5 years ago
- Views:
Transcription
1 Nonparameric Saisics, 2003, Vol. 15(2), pp NONPARAMETRIC ESTIMATION FOR MIDDLE-CENSORED DATA S. RAO JAMMALAMADAKA a, * and VASUDEVAN MANGALAM b a Deparmen of Saisics and Applied Probabiliy, Universiy of California, Sana Barbara, CA 93106, USA; b Deparmen of Mahemaics, Universii Brunei Darussalam, Brunei (Received 2 November 1999; In final form 22 Augus 2000) This paper provides he self-consisen esimaor (SCE) and he nonparameric maximum likelihood esimaor (NPMLE) for middle-censored daa, in which a daa value becomes unobservable if i falls wihin a random inerval. We provide an algorihm o find he SCE and show ha he NPMLE saisfies he self-consisency equaion. We find a sufficien condiion for he SCE o be concenraed on he uncensored observaions. In addiion, we find sufficien condiions for he consisency of he SCE and prove ha consisency holds for he special case when one of he ends is a consan. Some simulaion resuls and an illusraive example, using Danish melanoma daa se, are provided. Keywords: Survival funcion; Middle-censoring; Self-consisency; Nonparameric maximum likelihood esimaion AMS 1991 Subjec Classificaions: Primary: 62G05, 62G30; Secondary: 62G99 1 INTRODUCTION Esimaion of he unknown disribuion of a random variable is of fundamenal imporance in saisics. In areas such as reliabiliy, biomery, general medical follow-up sudies and clinical rials, he disribuion funcion of he underlying lifeime or more specifically. he survival funcion is of paramoun ineres. In hese siuaions, he random variable of ineres is he lifeime and he observaions refer o imes of occurrence of an even such as deah due o a cerain cause under sudy, or imes for equipmen failure. When complee daa are available, he Empirical Disribuion Funcion (EDF) is used and i has many desirable properies. However, in many pracical siuaions, i is quie common o have incomplee daa, making he sandard empirical disribuion funcion (EDF) unavailable. Ofen, such incomplee observaion of he daa resuls from a random censoring mechanism. When observaions are censored o he righ, he produc limi esimaor due o Kaplan and Meier (1958) is used in place of he EDF and similar esimaors exis for he lef-censored case. Gehan (1965) and Turnbull (1974) and ohers considered doubly-censored daa (where boh lef and righ censoring occur simulaneously) and esimaors for he disribuion funcion have been developed. Groeneboom and Wellner * Corresponding auhor. ISSN prin; ISSN online # 2003 Taylor & Francis Ld DOI: =
2 254 S. R. JAMMALAMADAKA AND V. MANGALAM (1992) and Geskus and Groeneboom (1996) sudied he case of inerval-censored daa where one can only observe a censoring even and wheher he ime of he even of ineres, say deah, occurred before or afer he occurrence of he censoring even. Nonparameric Maximum Likelihood Esimaors (NPMLE) for he disribuion of ineres have been sudied by various auhors for all hese cases. A Self Consisen Esimaor (SCE) is usually obained by solving a se of equaions called he self-consisency equaions (see Efron, 1967; Tarpey and Flury, 1996), and under some condiions his coincides wih he NPMLE. Tsai and Crowley (1985) have shown ha many of hese cases can be unified by applying a generalized maximum likelihood principle. They also poin ou ha solving he self-consisency equaion is essenially equivalen o applying he EM algorihm for he corresponding missing daa problem. See Dempser and Laird (1977) and McLachlan and Krishnan (1997) on he EM algorihm. In his paper we consider an imporan variaion and generalizaion of censoring where a daa poin becomes unobservable if i fell inside a random inerval. When ha happens we observe a censorship indicaor and he inerval of censorship. We will refer o his as middle-censoring. Lef censoring, righ censoring and double censoring are special cases of his middle censorship by suiable choice of his censoring inerval, which can be infinie. Middle-censoring where a random middle par is missing appears a firs glance, as complemenary o he idea of double-censoring where he middle is wha is acually observed. However, if one considers hese wo schemes carefully along wih he resuling daa ses (see nex secion), hey urn ou o be quie disinc ideas. Before discussing he esimaor, we consider some siuaions where his ype of censoring may arise. In general, in any lifeime sudy, if he subjec is emporarily wihdrawn from he sudy we will have an inerval of censorship. I can be equipmen failure ha could occur during a period where observaion is no possible or is no being made. In biomedical sudies, he paien under observaion may be absen from sudy for a shor period during which ime he even of ineres may occur. As an example of double censoring, Turnbull (1974) refers o a sudy of African infan precociy by Leiderman e al. (1973), where esablishing he for infan developmen for a communiy in Kenya was he purpose. A sample of 65 children are considered and each child was esed monhly o see if (s)he had learned o accomplish cerain asks. The ime from birh o he learning ime was he variable of ineres. In heir analysis, double-censoring occurred due o lae arrivals (he child had already learned he skills before enering he sudy) and losses (he child had no acquired he skill by he end of ime sudy). We envisage a scenario where here are no lae enries or losses as such, bu during a fixed ime inerval (his fixed inerval is indeed, a random inerval relaive o he individual s lifeime) he observaion was no possible, such as he emporary closure of he clinic due o an oubreak of say, war. If some children, of varying ages, developed he skill during his ime, we do no observe he exac age of hese children a he ime of skill developmen, raher only he informaion ha he even of ineres occurred during a cerain ime inerval. These ideas can, of course, be exended o more general random ses of censorship such as union of inervals or even more complicaed ses bu we have no explored his in deail. In Secion 2, we derive he self-consisency equaion for he middle censored case and show ha he NPMLE indeed saisfies he self-consisency equaion. A simple example which shows how one compues he NPMLE is also given. In Secion 3 we explore condiions under which he self-consisen esimaor (SCE) is consisen and prove he consisency in an imporan special case. Secion 4 illusraes he SCE for middle-censored case for a simulaed daa se as well as for a real daa se on Melanoma survival, from Andersen e al. (1993). A compuer program which allows he compuaion of he SCE is available by wriing o he auhors.
3 2 SELF-CONSISTENCY AND THE NPMLE Le X i, i ¼ 1,..., n, be a sequence of independen idenically disribued (i.i.d.) random variables wih unknown disribuion F 0. Le Y i ¼ (L i, R i ) be a sequence of i.i.d. random vecors, independen of X s, wih unknown bivariae disribuion G such ha P(L i < R i ) ¼ 1. While X denoes he variable of ineres, Y represens he censoring mechanism. Using he noaion we observe d i ¼ I[X i =2 (L i, R i )], Z i ¼ X i when d i ¼ 1 i:e:, if X i =2 (L i, R i ) ¼ (L i, R i ) when d i ¼ 0 i:e:, if X i 2 (L i, R i ) Tha is, we eiher observe he original value X i, if here is no censoring or he inerval of censoring (L i, R i ) when here is censoring. In many censoring siuaions, if we were o ry o esimae he disribuion funcion via he EM algorihm he resuling equaion akes he form ^F() ¼ E ^F [E n()jz] as described by Tsai and Crowley (1985), where E n is he empirical disribuion funcion. This equaion was firs inroduced and referred o as self-consisency equaion by Efron (1967). In he middle censored case he SCE F n, saisfies he equaion F n () ¼ 1 X n d i I(X i ) þ d n i I(R i ) þ d F n () F n (L i ) i I[ 2 (L i,r i )] (1) F n (R i ) F n (L i ) i¼1 where d i ¼ 1 d i. (For he res of he paper we will follow he convenion ha x, for any variable or funcion x, indicaes 1 x). As in he case of doubly-censored daa, here is no explici closed form soluion o he equaion and has o be compued by he ieraive formula ^F (mþ1) () ¼ E ^F (m) [E n ()jz]: The convergence of he algorihm is assured by Theorem 2.1 of Tsai and Crowley (1985) provided ha he iniial esimaor gives posiive mass o all observed poins. See Remark 2.1 below regarding he choice of he iniial esimaor. For a general discussion on selfconsisency and is relaion o EM algorihm, see Tarpey and Flury (1996). Le F denoe he se of all disribuion funcions on he line. For F 2F he likelihood of he sample is given by L(F) ¼ Yn i¼1 [F(X i ) F(X i )] d i [F(R i ) F(L i )] 1 d i : Denoing by DF(x) ¼ F(x) F(x ), f(f) (1=n) log L(F) is given by X n ESTIMATION FOR MIDDLE-CENSORED DATA 255 f(f) ¼ 1 [d i log(df(x i )) þ d n i log [F(R i ) F(L i )]] i¼1 ¼ {I[x =2 (l, r)] log DF(x) þ I[x 2 (l, r)] log[f(r ) F(l)]} dp n (x, l, r)
4 256 S. R. JAMMALAMADAKA AND V. MANGALAM where P n is he empirical measure of {(X i, L i, R i ): 1 i n}. The maximizer of f is clearly he NPMLE. In he nex heorem, we show ha he NPMLE for middle censored daa saisfies he self-consisency equaion. Bu before ha we need he following lemma. LEMMA 1 Define F( ^ x) A (x) ¼ F(x) F() if F() > 0 ¼ 0 oherwise (2) where ^ x ¼ min(, x). Then K(x) ¼ F(x) þ ha (x) defines a class of disribuion funcions for h sufficienly close o zero. Proof Noe ha we need o show his only for F() > 0 since A 0 when F() ¼ 0. F( ^ x) K(x) ¼ (1 h)f(x) þ h F() is a convex combinaion of wo cdf s and hence is a cdf for 0 h < 1. For negaive h, wrie K ¼ F ha wih h > 0 so ha K(x) ¼ (1 þ h)f(x) h F( ^ x) : F() Clearly K( 1) ¼ 0 and K(1) ¼ 1. I is also righ-coninuous so i remains o show ha i is monoone. We check his separaely on ( 1, ] and [, 1). For x in ( 1, ], K(x) ¼ (1 þ h)f(x) h F(x) F() (1 F()) ¼ F(x) 1 h : F() This is clearly bounded by 1 and is non-negaive if h (F()=1 F()) and in his case, K is monoone non-decreasing. Similarly on [, 1), K(x) ¼ (1 þ h)f(x) h ¼ F(x) h(1 F(x)): Again his is bounded by 1 and is non-negaive so long as h (F(x)=1 F(x)) which is assured if h (F()=1 F()) since x. Monooniciy of K is clear. j THEOREM 1 The NPMLE saisfies he equaion F() ¼ 1 n X n i¼1 d i I(X i ) þ d i I(R i ) þ d F() F(L i ) i I[ 2 (L i, R i )] : (3) F(R i ) F(L i )
5 Proof If F maximizes f, hen he direcional derivaive of f owards A should be zero a F i.e., saisfies he equaion f(f þ ha ) f(f) 0 ¼ lim h!0 h log[df(x) þ hda (x)] log DF(x) ¼ I[x =2 (l, r)] lim h!0 h þi[x 2 (l, r)] lim h!0 log[f(r ) F(l) þ h(a (r ) A (l))] log [F(r ) F(l)] h dp n (x, l, r) ESTIMATION FOR MIDDLE-CENSORED DATA 257 as he inegral involved is really a finie sum and hence inerchanging of limi and inegraion is valid. When F() > 0, he firs of he wo limis inside he inegral is log[df(x) þ e] log DF(x) DA (x) lim ¼ DA (x) e!0 e DF(x) The second limi is similarly equal o A (r ) A (l) F(r ) F(l) F() ^ F(r ) F( ^ I) ¼ 1 F()[F(r ) F(l)] I(x ) ¼ 1: F() where F() ^ F(r ) sands for F() if < r and F(r ) oherwise. Thus we ge or 1 ¼ I[x =2 (l, r)] I(x ) þ I[x 2 (l, r)] F() F() ^ F(r ) F( ^ I) F()[F(r ) F(l)] dp n (x, l, r) F() ¼ I[x =2 (l, r)]i(x ) þ I(l < x < r ) þ I[x, 2 (l, r)]: F() F(l) dp n (x, l, r): F(r ) F(l) RHS of he above is same as RHS of (3). j I is a quesion of considerable ineres o ask if NPMLE will have all is mass on he uncensored observaions. The answer is yes, provided all censored inervals conain a leas one uncensored observaion. When here is a censoring inerval empy of uncensored observaions, clearly some mass has o be aached o ha inerval or he likelihood would be zero. Tha he weighs are concenraed on he uncensored observaions when all censoring inervals are non-empy is a consequence of he following proposiion. PROPOSITION 1 If each observed censored inerval (L i, R i ) conains a leas one uncensored observaion X j, j 6¼ i, hen any disribuion funcion ha saisfies (3) aaches all is mass on he uncensored observaions.
6 258 S. R. JAMMALAMADAKA AND V. MANGALAM Proof Le F be a disribuion saisfying (3) and le x 1, x 2,..., x m be he uncensored observaions. For any x le DF(x) ¼ F(x) F(x ) be he weigh F associaes o x. We need o show ha P m j¼1 DF(x j) ¼ 1. From (3) i follows ha DF(x j ) ¼ 1 n þ 1 n Summing (4) over all uncensored observaions, we ge X m X n i¼1 DF(x j ) (1 d i )I[x j 2 (L i, R i )] : (4) F(R i ) F(L i ) X n DF(x j ) ¼ m n þ 1 (1 d i ) P j I[x j 2 (L i, R i )]DF(x j ) (5) n F(R j¼1 i¼1 i ) F(L i ) For each censoring inerval (L i, R i ), le a i be he slack beween he mass associaed o he inerval and he sum of weighs of all uncensored observaions in he inerval. Then a i ¼ F(R i ) F(L i ) Xm I[x j 2 (L i, R i )]DF(x j ) (6) j¼1 and a i s are all non-negaive. From (5) and (6), i follows ha 1 X a i ¼ 1 1 X a i n F(R i ) F(L i ) or X ai ¼ 1 X a i (7) n F(R i ) F(L i ) where he sum is over all censored observaions. As every inerval conains a leas one uncensored observaion, i follows from (4) ha F(R i ) F(L i ) (1=n) and hence (7) implies ha a i a i ¼ (8) n(f(r i ) F(L i )) for each i. Now if here exiss i such ha a i > 0, (F(R i ) F(L i ))(1=n) þ a i > (1=n) conradicing (8). j We have now proved ha he NPMLE will have all is mass on he uncensored observaions excep when i so happens ha a censored inerval conains no uncensored observaion. If his happens, we are in a siuaion similar o ha of righ censored daa where he larges observaion is censored. While in he righ censored case he exra mass is usually lef unassigned, for middle-censored daa here is a naural way of handling his. When a censored inerval conains no uncensored poins, we le he mass ha corresponds o ha inerval be assigned o is midpoin. Thus our iniial esimaor may give equal mass o all uncensored observaions and o he midpoins of hose finie censored inervals ha conain no uncensored observaions. If an infinie censoring inerval happens o be empy of uncensored observaions, one can hen assign he mass o any arbirary poin inside his inerval for he esimaor o have a maximum. Consider he following example where n ¼ 5 and z 1 ¼ 2, z 2 ¼ 4, z 3 ¼ 6, z 4 ¼ (1, 5) and z 5 ¼ (3, 7). Le p 1, p 2, p 3 be he masses o be assigned o z 1, z 2, z 3 respecively. The likelihood funcion is given by p 1 p 2 p 3 ( p 1 þ p 2 )( p 2 þ p 3 )
7 ESTIMATION FOR MIDDLE-CENSORED DATA 259 and, as p i s add up o 1 and he roles of p 1 and p 3 are inerchangeable, we can simplify he problem o ha of maximizing (x 2 )(1 p 2x)(1 x) 2 wih p 1 ¼ p 3 ¼ x and p 2 ¼ 1 2x. The soluion, hen, is given by x ¼ (5 ffiffiffi p 5 )=10 so ha p1 ¼ p 3 ¼ (5 ffiffi p 5 )=10 and p2 ¼ 1= ffiffi 5 is he soluion o he NPMLE. In his example he ieraions of he self-consisency equaion rapidly converged o he NPMLE. The SCE, being a resul of convergence of he EM algorihm, provides a local maximum of he likelihood equaion [see, for example, Mykland and Ren, 1996] and may no coincide wih he NPMLE. Examples of cases when an SCE is no he NPMLE can be consruced by considering siuaions where wo empy censoring inervals overlap. For insance, if we have 1, 2, (3, 6), (4, 7) as he daa, we could assign 0.25 mass o 1, 2, 4.5 and 5.5 o ge an SCE. The NPMLE will assign 0.25 each on 1 and 2, bu assign 0.5 on some poin, say 5, on he overlap area (4, 6). Boh esimaors are self-consisen, bu he laer has higher likelihood. This happens whenever here are empy, overlapping inervals. In he nex secion we shall show he srong consisency of self-consisen esimaors for cerain cases. This will demonsrae ha SCE and NPMLE are, a leas for hese special cases, asympoically equivalen and hence will be approximaely he same for large samples. 3 CONSISTENCY OF SELF-CONSISTENT ESTIMATORS Define P and Q, sub-disribuion funcions on R and R 2 respecively, by and heir empirical versions P n and Q n by P() ¼ P(X, d ¼ 1) Q(l, r) ¼ P(L l, R r, d ¼ 0) P n () ¼ 1 n Q n (l, r) ¼ 1 n X n i¼1 X n i¼1 I(X i, d i ¼ 1) I(L i l, R i r, d i ¼ 0): Then by Glivenko-Canelli Lemma, i follows ha P n and Q n converge almos surely o P and Q respecively and he convergence in each case is uniform on he respecive domain. Also, (1) can be wrien in erms of P n and Q n as follows: Fn () ^ F n (r ) F n ( ^ l) F n () ¼ P n () þ dq n (l, r) (9) F n (r ) F n (l) By Helly s Theorem, 9 a subsequence n k and a non-decreasing funcion F bounded by 0 and 1 such ha on a se of probabiliy 1, F nk ()!F() for each. PROPOSITION 2 If {j n } is a sequence of funcions on R 2 which converge uniformly o bounded coninuous funcion j, hen j n (l, r)dq n (l, r)! j(l, r) dq(l, r):
8 260 S. R. JAMMALAMADAKA AND V. MANGALAM Proof Noe ha, j n (l, r)dq n (l, r) j n (l, r)dq(l, r) (j n (l, r) j(l, r)) dq n (l, r) þ j(l, r)dq n (l, r) j(l, r) dq(l, r) kj n jk dq n þ j(l, r)dq n (l, r) j(l, r) dq(l, r) where kk represens he supremum norm. Now he firs erm on he RHS of he inequaliy goes o zero since Ð dq n ¼ 1 while he second erm goes o zero because he sequence of empirical measures Q n converge o Q weakly and j is a bounded coninuous funcion. j LEMMA 2 Any subsequenial limi F of F n will saisfy he equaion F() ^ F(r ) F( ^ l) F() ¼ P() þ dq(l, r) (10) F(r ) F(l) Proof For a fixed, aking limis in (9) hrough he subsequence n k as k!1and using Proposiion 2 wih j n (l, r) ¼ F n() ^ F n (r ) F n ( ^ l), F n (r ) F n (l) he resul follows. j When P and Q are wrien in erms of F 0 and G, (10) is equivalen o F() F(l) F() F 0 () ¼ F(r ) F(l) (F 0(r) F 0 (l)) þ F 0 (l) F 0 () dg(l, r): (11) l<<r From (11), i follows ha F(1) ¼ 1 and F( 1) ¼ 0. Noe ha if F ¼ F 0, (11) is auomaically saisfied. If we were able o show ha (11) has a unique soluion, hen i follows ha F 0 is he only limi poin of {F n }. Then we will have ha on a se of probabiliy 1, F n (x)! F 0 (x) for each x and by coninuiy of F 0 uniformiy of he almos sure convergence follows. A necessary condiion for consisency is wha we call idenifiabiliy. Le A() ¼ P(L < < R). The condiion is ha A be no idenically 1 on any inerval [a, b], a b for which F 0 (b) > F 0 (a ). Observe ha if A 1 on any inerval where F 0 has a posiive mass, hen censoring occurs wih probabiliy 1 on such an inerval. As a consequence, here will be no observaions on his inerval and ha prevens us from disinguishing any wo disribuions which are idenical ouside [a, b] bu differing only on [a, b]. This condiion will be referred o as he idenifiabiliy condiion and is a requiremen for consisen esimaion of F 0. LEMMA 3 Le h ¼ (F F 0 ) and g() ¼ c(l, r)i(l < < r)dg(l, r) R 2
9 ESTIMATION FOR MIDDLE-CENSORED DATA 261 where c(l, r) ¼ (h(r) h(l)=f(r ) F(l)). Then Adh ¼ gdf (12) Proof From (11) we ge h() ¼ ¼ l<<r l<<r h(r) h(l) (F() F(l)) þ h(l) h() dg(l, r) F(r ) F(l) [(F() F(l))c(l, r) þ h(l)]dg(l, r) þ h()a(): So, where h() A() ¼ l<<r [(F() F(l))c(l, r) þ h(l)]dg(l, r) ¼ F()g() þ C() (13) C() ¼ l<<r Differeniaing boh sides of (13) w.r.., we ge [h(l) F(l)c(l, r)]dg(l, r): h()da() A()dh() ¼ g()df() þ F()dg() þ dc(): To show ha (12) holds, clearly i is sufficien o show ha F()dg() þ dc() h()da() ¼ 0 (14) If B() ¼ Ð l<<r H(l, r)dg(l, r) for some funcion H, hen i can be shown ha db() ¼ ( Ð 1 H(, r)df RjL (rj))df L () ( Ð 1 H(l, )df LjR(lj))dF R () where F RjL ( j) is he condiional disribuion funcion of R given L ¼. Hence, applying his o g and C, 1 LHS of(14) ¼ F() c(, r)df RjL (rj) df L () F() c(l, )df LjR (lj) df R () 1 1 þ (h() F()c(, r))df RjL (rj) df L () (h(l) F(l)c(l, ))df LjR (lj) df R () h()da() 1 ¼ df R () 1 h()df LjR (lj) þ df L () 1 h()df RjL (rj) h()da() ¼ h()(df L () df R ()) h()da() ¼ 0 because Ð 1 df LjR(lj) ¼ Ð 1 df RjL (rj) ¼ 1 and A() ¼ F L () F R (): j Thus, if he only funcion h saisfying (12) is he zero funcion, hen we would have proved he srong consisency of he SCE. We have no ye been able o show ha his (12) has a unique soluion in he general case, bu we give below a proof for he special case when one of he end poins of he censoring inerval is degenerae. Alhough he resul is saed for he case when L is degenerae (for insance, censoring if i occurs, sars on a cerain birhday of he individual), he proof works equally well when R is degenerae.
10 262 S. R. JAMMALAMADAKA AND V. MANGALAM THEOREM 2 Assume ha F 0 and F R are coninuous and L l 0. Assume he idenifiabiliy condiion is saisfied. Then he only funcion F ha saisfies 11Þ is F 0 and hence he SCE is uniformly srongly consisen. Proof In his special case (13) becomes h() A() ¼ I( > l 0 ) 1 [(F() F(l 0 ))c(l 0, r) þ h(l 0 )]df R (r) (15) As A() ¼ 1 P[ 2 (l 0, R)] ¼ 1 I( > l 0 ) F R (), A() ¼ 1 for all l 0 and A() ¼ F R () for all > l 0 ; hence from (15), h() ¼ 0 for all l 0. In paricular, h(l 0 ) ¼ 0. Thus (15) becomes 1 F() F(l 0 ) h() A() ¼ I( > l 0 ) F(r ) F(l 0 ) h(r)df R(r): Similarly, (12) holds wih 1 h(r) g() ¼ I( > l 0 ) F(r ) F(l 0 ) df R(r): (16) Noe ha from he assumpions of he heorem i follows ha F, h and g are coninuous on (l 0, 1). We aim o show ha h 0on(l 0, 1). Assuming 9 0 > l 0 such ha h( 0 ) > 0, we will arrive a a conradicion. The proof is similar if h( 0 ) < 0. j As h(l 0 ) ¼ h(1) ¼ 0, 9 1 < 2 such ha 1 l 0, 2 1, h( 1 ) ¼ h( 2 ) ¼ 0 and h() > 0on ( 1, 2 ). From (16), on (l 0, 1), g() ¼ Ð 1 c(r)df R (r) where c(r) ¼ (h(r)=f(r ) F(l 0 )). CLAIM 1 g( 1 ) 0. Suppose no. Then 9 such ha g > 0on( 1, ). We shall now show ha A() > 0on ( 1, ). If 9 2 ( 1, ) such ha A() ¼ 0, hen A() ¼ 0 for all 2 (l 0,) so ha by he idenifiabiliy condiion, df 0 () ¼ 0 for all 2 (l 0, ). From (12) df() ¼ 0on( 1, ); so dh() ¼ df() df 0 () ¼ 0on( 1, ) which implies h 0 here, conrary o our assumpion. From (12), dh() ¼ ( g()=1 A())dF(), so Ð 1 ( g()=1 A())dF() ¼ h( ) h( 1 ) ¼ h( ). Now, LHS 0 conradicing he fac ha h( ) > 0. This proves Claim 1. CLAIM 2 g( 2 ) 0. Suppose no. Then 9 2 ( 0, 2 ) such ha g < 0on(, 2 ). Similar o he previous siuaion, we have A() > 0on(, 2 ). As earlier, h( 2 ) h( ) ¼ Ð 2 ( g()=1 A())dF() 0, implying h( ) 0 which is conradicion. Thus Claim 2 is proved. On ( 1, 2 ), dg() ¼ c()df R () ¼ ( h()df R ()=F( ) F(l 0 )) 0, so g is decreasing here. Thus from Claim 1 and Claim 2 i follows ha g 0on( 1, 2 ). (Noe ha he argumen Ð goes hrough even if 2 ¼1). From (12), Adh 0 on ( 1, 2 ). As g() ¼ 1 c(r)df R (r), F R is a consan ( 1, 2 ) and hence A is a consan on (1, 2 ). If c > 0, h 0on( 1, 2 ) which is a conradicion. If c ¼ 0, A 0on( 1, 2 ) and hence by he idenifiabiliy condiion F 0 is consan on ( 1, 2 ). As h( 1 ) ¼ h( 2 ) ¼ 0, F( 1 ) ¼ F( 2 ), so F is a consan on ( 1, 2 ). So h is a consan on ( 1, 2 ), which means h 0on( 1, 2 ). This is a conradicion. 4 ILLUSTRATIVE EXAMPLES A simulaion sudy was performed o measure he performance of self-consisen esimaors where an exponenial (mean ¼ 10) random variable was middle-censored by random inervals
11 ESTIMATION FOR MIDDLE-CENSORED DATA 263 FIGURE 1 The EDF and SCE for he simulaed exponenial daa. wih lef end poins being exponenial (mean ¼ 5) and inerval widhs being an independen exponenial (mean ¼ 5). A sample of size 100 was aken and his resuled in 22 of hem being censored. Figure 1 shows he SCE along wih he original exponenial disribuion funcion F 0. The maximum disance kf n F 0 k is which is very good compared o he Kolmogorov Smirnov disance, namely he maximum disance of he EDF. E n of he uncensored daa from he rue disribuion, which is ke n F 0 k¼0:0715. The auhors also ried ou various oher survival disribuions such as gamma and Weibull ha were censored by inervals whose lef ends were disribued as exponenial, gamma, Weibull or uniform and inerval widh was a posiive random variable or a consan. In all hese cases, he resuling esimaors for middle censoring were in very close agreemen wih he EDF of he original uncensored daa. I is clear ha he amoun of censoring in any of hese cases, is approximaely P(L X R). Finally we applied our echniques o an acual daa se on melanoma survival colleced a Odense Universiy Hospial, Denmark [see Andersen e al., 1993]. The sample conains 205 daa poins, ranging from 10 o The daa were censored by a random inerval whose lef end was an exponenial random variable wih mean 2000 and widh was exponenial wih mean Over 23% of daa were censored resuling in 157 uncensored observaions. The SCE F n is given in Figure 2 while he EDF E n of he survival daa is in Figure 3. They are shown super-imposed in Figure 4, o see how close hey are. Indeed, he maximum disance kf n E n k beween hem is while he maximum relaive disance k((f n E n )=E n )k urns ou o be sill a small FIGURE 2 SCE for he censored melanoma survival daa.
12 264 S. R. JAMMALAMADAKA AND V. MANGALAM FIGURE 3 EDF for he uncensored melanoma survival daa. FIGURE 4 EDF and SCE superimposed. Acknowledgemens We would like o hank an anonymous referee whose persisence led o a much more readable paper. References Andersen, P. K., Borgan, O., Gill, R. D. and Keiding, N. (1993). Saisical Models Based on Couning Processes. Springer-Verlag, New York. Dempser, A. P., Laird, N. M. and Rubin, D. B. (1977). Maximum likelihood from incomplee daa via he EM algorihm. J. Roy. Sais. Soc. Ser., B39, 1 38 (wih discussion). Efron, B. (1967). The wo-sample problem wih censored daa. Proc. Fifh Berkeley Symp. Mah. Sa. Probab., Vol. 4. Universiy of California Press, Berkeley, pp Gehan, E. A. (1965). A generalized wo-sample Wilcoxon es for doubly censored daa. Biomerika, 52, Geskus, R. B. and Groeneboom, P. (1996). Asympoically opimal esimaion of smooh funcionals for inerval censoring. J. Sais. Neerlandica, 50, Groeneboom, P. and Wellner, J. A. (1992). Informaion Bounds and Non-parameric Maximum Likelihood Esimaion. DMV Seminar, Vol. 19. Birkhäuser Verlag, Basel. Gu, M. G. and Zhang, C.-H. (1993). Asympoic properies of self-consisen esimaors based on doubly censored daa. Ann. Sais., 21, Kaplan, E. L. and Meier, P. (1958). Nonparameric esimaion from incomplee observaions. J. Amer. Sais. Assoc., 53, Leiderman, P. H., Babu, D., Kagia, J., Kraemer, H. C. and Leiderman, G. F. (1973). African infan precociy and some social influences during he firs year. Naure, 242, McLachlan, G. J. and Krishnan, T. (1997). The EM Algorihm and Exensions. John Wiley and Sons, Inc., New York.
13 ESTIMATION FOR MIDDLE-CENSORED DATA 265 Mykland, P. A. and Ren, J. (1996). Algorihms for compuing self-consisen and maximum likelihood esimaors wih doubly censored daa. Ann. Sais., 24, Tarpey, T. and Flury, B. (1996). Self-consisency: A fundamenal concep in saisics. Saisical Science, 11, Tsai, W. Y. and Crowley, J. (1985). A large sample sudy of generalized maximum likelihood esimaors from incomplee daa via self-consisency. Ann. Sais., 13, Turnbull, B. W. (1974). Nonparameric esimaion of a survivorship funcion wih doubly censored daa. J. Amer. Sais. Assoc., 69,
14
The Asymptotic Behavior of Nonoscillatory Solutions of Some Nonlinear Dynamic Equations on Time Scales
Advances in Dynamical Sysems and Applicaions. ISSN 0973-5321 Volume 1 Number 1 (2006, pp. 103 112 c Research India Publicaions hp://www.ripublicaion.com/adsa.hm The Asympoic Behavior of Nonoscillaory Soluions
More informationChapter 2. Models, Censoring, and Likelihood for Failure-Time Data
Chaper 2 Models, Censoring, and Likelihood for Failure-Time Daa William Q. Meeker and Luis A. Escobar Iowa Sae Universiy and Louisiana Sae Universiy Copyrigh 1998-2008 W. Q. Meeker and L. A. Escobar. Based
More informationExpert Advice for Amateurs
Exper Advice for Amaeurs Ernes K. Lai Online Appendix - Exisence of Equilibria The analysis in his secion is performed under more general payoff funcions. Wihou aking an explici form, he payoffs of he
More informationSolutions from Chapter 9.1 and 9.2
Soluions from Chaper 9 and 92 Secion 9 Problem # This basically boils down o an exercise in he chain rule from calculus We are looking for soluions of he form: u( x) = f( k x c) where k x R 3 and k is
More informationVehicle Arrival Models : Headway
Chaper 12 Vehicle Arrival Models : Headway 12.1 Inroducion Modelling arrival of vehicle a secion of road is an imporan sep in raffic flow modelling. I has imporan applicaion in raffic flow simulaion where
More informationOptimality Conditions for Unconstrained Problems
62 CHAPTER 6 Opimaliy Condiions for Unconsrained Problems 1 Unconsrained Opimizaion 11 Exisence Consider he problem of minimizing he funcion f : R n R where f is coninuous on all of R n : P min f(x) x
More informationExponential Weighted Moving Average (EWMA) Chart Under The Assumption of Moderateness And Its 3 Control Limits
DOI: 0.545/mjis.07.5009 Exponenial Weighed Moving Average (EWMA) Char Under The Assumpion of Moderaeness And Is 3 Conrol Limis KALPESH S TAILOR Assisan Professor, Deparmen of Saisics, M. K. Bhavnagar Universiy,
More informationApproximate Self Consistency for Middle-Censored Data
Approximate Self Consistency for Middle-Censored Data by S. Rao Jammalamadaka Department of Statistics & Applied Probability, University of California, Santa Barbara, CA 93106, USA. and Srikanth K. Iyer
More informationLecture 33: November 29
36-705: Inermediae Saisics Fall 2017 Lecurer: Siva Balakrishnan Lecure 33: November 29 Today we will coninue discussing he boosrap, and hen ry o undersand why i works in a simple case. In he las lecure
More information2. Nonlinear Conservation Law Equations
. Nonlinear Conservaion Law Equaions One of he clear lessons learned over recen years in sudying nonlinear parial differenial equaions is ha i is generally no wise o ry o aack a general class of nonlinear
More informationChapter 2. First Order Scalar Equations
Chaper. Firs Order Scalar Equaions We sar our sudy of differenial equaions in he same way he pioneers in his field did. We show paricular echniques o solve paricular ypes of firs order differenial equaions.
More informationT L. t=1. Proof of Lemma 1. Using the marginal cost accounting in Equation(4) and standard arguments. t )+Π RB. t )+K 1(Q RB
Elecronic Companion EC.1. Proofs of Technical Lemmas and Theorems LEMMA 1. Le C(RB) be he oal cos incurred by he RB policy. Then we have, T L E[C(RB)] 3 E[Z RB ]. (EC.1) Proof of Lemma 1. Using he marginal
More information23.2. Representing Periodic Functions by Fourier Series. Introduction. Prerequisites. Learning Outcomes
Represening Periodic Funcions by Fourier Series 3. Inroducion In his Secion we show how a periodic funcion can be expressed as a series of sines and cosines. We begin by obaining some sandard inegrals
More informationRepresentation of Stochastic Process by Means of Stochastic Integrals
Inernaional Journal of Mahemaics Research. ISSN 0976-5840 Volume 5, Number 4 (2013), pp. 385-397 Inernaional Research Publicaion House hp://www.irphouse.com Represenaion of Sochasic Process by Means of
More informationRobust estimation based on the first- and third-moment restrictions of the power transformation model
h Inernaional Congress on Modelling and Simulaion, Adelaide, Ausralia, 6 December 3 www.mssanz.org.au/modsim3 Robus esimaion based on he firs- and hird-momen resricions of he power ransformaion Nawaa,
More informationInternational Journal of Scientific & Engineering Research, Volume 4, Issue 10, October ISSN
Inernaional Journal of Scienific & Engineering Research, Volume 4, Issue 10, Ocober-2013 900 FUZZY MEAN RESIDUAL LIFE ORDERING OF FUZZY RANDOM VARIABLES J. EARNEST LAZARUS PIRIYAKUMAR 1, A. YAMUNA 2 1.
More informationSection 3.5 Nonhomogeneous Equations; Method of Undetermined Coefficients
Secion 3.5 Nonhomogeneous Equaions; Mehod of Undeermined Coefficiens Key Terms/Ideas: Linear Differenial operaor Nonlinear operaor Second order homogeneous DE Second order nonhomogeneous DE Soluion o homogeneous
More information4 Sequences of measurable functions
4 Sequences of measurable funcions 1. Le (Ω, A, µ) be a measure space (complee, afer a possible applicaion of he compleion heorem). In his chaper we invesigae relaions beween various (nonequivalen) convergences
More informationMATH 4330/5330, Fourier Analysis Section 6, Proof of Fourier s Theorem for Pointwise Convergence
MATH 433/533, Fourier Analysis Secion 6, Proof of Fourier s Theorem for Poinwise Convergence Firs, some commens abou inegraing periodic funcions. If g is a periodic funcion, g(x + ) g(x) for all real x,
More informationIntroduction to Probability and Statistics Slides 4 Chapter 4
Inroducion o Probabiliy and Saisics Slides 4 Chaper 4 Ammar M. Sarhan, asarhan@mahsa.dal.ca Deparmen of Mahemaics and Saisics, Dalhousie Universiy Fall Semeser 8 Dr. Ammar Sarhan Chaper 4 Coninuous Random
More informationMatrix Versions of Some Refinements of the Arithmetic-Geometric Mean Inequality
Marix Versions of Some Refinemens of he Arihmeic-Geomeric Mean Inequaliy Bao Qi Feng and Andrew Tonge Absrac. We esablish marix versions of refinemens due o Alzer ], Carwrigh and Field 4], and Mercer 5]
More informationCHERNOFF DISTANCE AND AFFINITY FOR TRUNCATED DISTRIBUTIONS *
haper 5 HERNOFF DISTANE AND AFFINITY FOR TRUNATED DISTRIBUTIONS * 5. Inroducion In he case of disribuions ha saisfy he regulariy condiions, he ramer- Rao inequaliy holds and he maximum likelihood esimaor
More informationFinish reading Chapter 2 of Spivak, rereading earlier sections as necessary. handout and fill in some missing details!
MAT 257, Handou 6: Ocober 7-2, 20. I. Assignmen. Finish reading Chaper 2 of Spiva, rereading earlier secions as necessary. handou and fill in some missing deails! II. Higher derivaives. Also, read his
More information15. Vector Valued Functions
1. Vecor Valued Funcions Up o his poin, we have presened vecors wih consan componens, for example, 1, and,,4. However, we can allow he componens of a vecor o be funcions of a common variable. For example,
More informationKEY. Math 334 Midterm I Fall 2008 sections 001 and 003 Instructor: Scott Glasgow
1 KEY Mah 4 Miderm I Fall 8 secions 1 and Insrucor: Sco Glasgow Please do NOT wrie on his eam. No credi will be given for such work. Raher wrie in a blue book, or on our own paper, preferabl engineering
More informationEXERCISES FOR SECTION 1.5
1.5 Exisence and Uniqueness of Soluions 43 20. 1 v c 21. 1 v c 1 2 4 6 8 10 1 2 2 4 6 8 10 Graph of approximae soluion obained using Euler s mehod wih = 0.1. Graph of approximae soluion obained using Euler
More information6.2 Transforms of Derivatives and Integrals.
SEC. 6.2 Transforms of Derivaives and Inegrals. ODEs 2 3 33 39 23. Change of scale. If l( f ()) F(s) and c is any 33 45 APPLICATION OF s-shifting posiive consan, show ha l( f (c)) F(s>c)>c (Hin: In Probs.
More informationHamilton- J acobi Equation: Weak S olution We continue the study of the Hamilton-Jacobi equation:
M ah 5 7 Fall 9 L ecure O c. 4, 9 ) Hamilon- J acobi Equaion: Weak S oluion We coninue he sudy of he Hamilon-Jacobi equaion: We have shown ha u + H D u) = R n, ) ; u = g R n { = }. ). In general we canno
More informationt 2 B F x,t n dsdt t u x,t dxdt
Evoluion Equaions For 0, fixed, le U U0, where U denoes a bounded open se in R n.suppose ha U is filled wih a maerial in which a conaminan is being ranspored by various means including diffusion and convecion.
More informationHeavy Tails of Discounted Aggregate Claims in the Continuous-time Renewal Model
Heavy Tails of Discouned Aggregae Claims in he Coninuous-ime Renewal Model Qihe Tang Deparmen of Saisics and Acuarial Science The Universiy of Iowa 24 Schae er Hall, Iowa Ciy, IA 52242, USA E-mail: qang@sa.uiowa.edu
More informationPOSITIVE SOLUTIONS OF NEUTRAL DELAY DIFFERENTIAL EQUATION
Novi Sad J. Mah. Vol. 32, No. 2, 2002, 95-108 95 POSITIVE SOLUTIONS OF NEUTRAL DELAY DIFFERENTIAL EQUATION Hajnalka Péics 1, János Karsai 2 Absrac. We consider he scalar nonauonomous neural delay differenial
More informationMODULE 3 FUNCTION OF A RANDOM VARIABLE AND ITS DISTRIBUTION LECTURES PROBABILITY DISTRIBUTION OF A FUNCTION OF A RANDOM VARIABLE
Topics MODULE 3 FUNCTION OF A RANDOM VARIABLE AND ITS DISTRIBUTION LECTURES 2-6 3. FUNCTION OF A RANDOM VARIABLE 3.2 PROBABILITY DISTRIBUTION OF A FUNCTION OF A RANDOM VARIABLE 3.3 EXPECTATION AND MOMENTS
More informationR t. C t P t. + u t. C t = αp t + βr t + v t. + β + w t
Exercise 7 C P = α + β R P + u C = αp + βr + v (a) (b) C R = α P R + β + w (c) Assumpions abou he disurbances u, v, w : Classical assumions on he disurbance of one of he equaions, eg. on (b): E(v v s P,
More information) were both constant and we brought them from under the integral.
YIELD-PER-RECRUIT (coninued The yield-per-recrui model applies o a cohor, bu we saw in he Age Disribuions lecure ha he properies of a cohor do no apply in general o a collecion of cohors, which is wha
More informationSolutions to Assignment 1
MA 2326 Differenial Equaions Insrucor: Peronela Radu Friday, February 8, 203 Soluions o Assignmen. Find he general soluions of he following ODEs: (a) 2 x = an x Soluion: I is a separable equaion as we
More informationInventory Analysis and Management. Multi-Period Stochastic Models: Optimality of (s, S) Policy for K-Convex Objective Functions
Muli-Period Sochasic Models: Opimali of (s, S) Polic for -Convex Objecive Funcions Consider a seing similar o he N-sage newsvendor problem excep ha now here is a fixed re-ordering cos (> 0) for each (re-)order.
More informationt is a basis for the solution space to this system, then the matrix having these solutions as columns, t x 1 t, x 2 t,... x n t x 2 t...
Mah 228- Fri Mar 24 5.6 Marix exponenials and linear sysems: The analogy beween firs order sysems of linear differenial equaions (Chaper 5) and scalar linear differenial equaions (Chaper ) is much sronger
More informationFinal Spring 2007
.615 Final Spring 7 Overview The purpose of he final exam is o calculae he MHD β limi in a high-bea oroidal okamak agains he dangerous n = 1 exernal ballooning-kink mode. Effecively, his corresponds o
More informationLecture 10: The Poincaré Inequality in Euclidean space
Deparmens of Mahemaics Monana Sae Universiy Fall 215 Prof. Kevin Wildrick n inroducion o non-smooh analysis and geomery Lecure 1: The Poincaré Inequaliy in Euclidean space 1. Wha is he Poincaré inequaliy?
More informationGMM - Generalized Method of Moments
GMM - Generalized Mehod of Momens Conens GMM esimaion, shor inroducion 2 GMM inuiion: Maching momens 2 3 General overview of GMM esimaion. 3 3. Weighing marix...........................................
More informationA new flexible Weibull distribution
Communicaions for Saisical Applicaions and Mehods 2016, Vol. 23, No. 5, 399 409 hp://dx.doi.org/10.5351/csam.2016.23.5.399 Prin ISSN 2287-7843 / Online ISSN 2383-4757 A new flexible Weibull disribuion
More informationAn Introduction to Malliavin calculus and its applications
An Inroducion o Malliavin calculus and is applicaions Lecure 5: Smoohness of he densiy and Hörmander s heorem David Nualar Deparmen of Mahemaics Kansas Universiy Universiy of Wyoming Summer School 214
More informationSemi-Competing Risks on A Trivariate Weibull Survival Model
Semi-Compeing Risks on A Trivariae Weibull Survival Model Jenq-Daw Lee Graduae Insiue of Poliical Economy Naional Cheng Kung Universiy Tainan Taiwan 70101 ROC Cheng K. Lee Loss Forecasing Home Loans &
More informationThe Strong Law of Large Numbers
Lecure 9 The Srong Law of Large Numbers Reading: Grimme-Sirzaker 7.2; David Williams Probabiliy wih Maringales 7.2 Furher reading: Grimme-Sirzaker 7.1, 7.3-7.5 Wih he Convergence Theorem (Theorem 54) and
More informationSPECTRAL EVOLUTION OF A ONE PARAMETER EXTENSION OF A REAL SYMMETRIC TOEPLITZ MATRIX* William F. Trench. SIAM J. Matrix Anal. Appl. 11 (1990),
SPECTRAL EVOLUTION OF A ONE PARAMETER EXTENSION OF A REAL SYMMETRIC TOEPLITZ MATRIX* William F Trench SIAM J Marix Anal Appl 11 (1990), 601-611 Absrac Le T n = ( i j ) n i,j=1 (n 3) be a real symmeric
More informationSupplement for Stochastic Convex Optimization: Faster Local Growth Implies Faster Global Convergence
Supplemen for Sochasic Convex Opimizaion: Faser Local Growh Implies Faser Global Convergence Yi Xu Qihang Lin ianbao Yang Proof of heorem heorem Suppose Assumpion holds and F (w) obeys he LGC (6) Given
More informationarxiv: v1 [math.fa] 9 Dec 2018
AN INVERSE FUNCTION THEOREM CONVERSE arxiv:1812.03561v1 [mah.fa] 9 Dec 2018 JIMMIE LAWSON Absrac. We esablish he following converse of he well-known inverse funcion heorem. Le g : U V and f : V U be inverse
More informationChapter 6. Systems of First Order Linear Differential Equations
Chaper 6 Sysems of Firs Order Linear Differenial Equaions We will only discuss firs order sysems However higher order sysems may be made ino firs order sysems by a rick shown below We will have a sligh
More information12: AUTOREGRESSIVE AND MOVING AVERAGE PROCESSES IN DISCRETE TIME. Σ j =
1: AUTOREGRESSIVE AND MOVING AVERAGE PROCESSES IN DISCRETE TIME Moving Averages Recall ha a whie noise process is a series { } = having variance σ. The whie noise process has specral densiy f (λ) = of
More informationProduct Integration. Richard D. Gill. Mathematical Institute, University of Utrecht, Netherlands EURANDOM, Eindhoven, Netherlands August 9, 2001
Produc Inegraion Richard D. Gill Mahemaical Insiue, Universiy of Urech, Neherlands EURANDOM, Eindhoven, Neherlands Augus 9, 21 Absrac This is a brief survey of produc-inegraion for biosaisicians. 1 Produc-Inegraion
More informationOscillation of an Euler Cauchy Dynamic Equation S. Huff, G. Olumolode, N. Pennington, and A. Peterson
PROCEEDINGS OF THE FOURTH INTERNATIONAL CONFERENCE ON DYNAMICAL SYSTEMS AND DIFFERENTIAL EQUATIONS May 4 7, 00, Wilmingon, NC, USA pp 0 Oscillaion of an Euler Cauchy Dynamic Equaion S Huff, G Olumolode,
More informationProperties Of Solutions To A Generalized Liénard Equation With Forcing Term
Applied Mahemaics E-Noes, 8(28), 4-44 c ISSN 67-25 Available free a mirror sies of hp://www.mah.nhu.edu.w/ amen/ Properies Of Soluions To A Generalized Liénard Equaion Wih Forcing Term Allan Kroopnick
More informationarxiv: v1 [math.pr] 19 Feb 2011
A NOTE ON FELLER SEMIGROUPS AND RESOLVENTS VADIM KOSTRYKIN, JÜRGEN POTTHOFF, AND ROBERT SCHRADER ABSTRACT. Various equivalen condiions for a semigroup or a resolven generaed by a Markov process o be of
More informationChallenge Problems. DIS 203 and 210. March 6, (e 2) k. k(k + 2). k=1. f(x) = k(k + 2) = 1 x k
Challenge Problems DIS 03 and 0 March 6, 05 Choose one of he following problems, and work on i in your group. Your goal is o convince me ha your answer is correc. Even if your answer isn compleely correc,
More informationIMPLICIT AND INVERSE FUNCTION THEOREMS PAUL SCHRIMPF 1 OCTOBER 25, 2013
IMPLICI AND INVERSE FUNCION HEOREMS PAUL SCHRIMPF 1 OCOBER 25, 213 UNIVERSIY OF BRIISH COLUMBIA ECONOMICS 526 We have exensively sudied how o solve sysems of linear equaions. We know how o check wheher
More informationA Note on Superlinear Ambrosetti-Prodi Type Problem in a Ball
A Noe on Superlinear Ambrosei-Prodi Type Problem in a Ball by P. N. Srikanh 1, Sanjiban Sanra 2 Absrac Using a careful analysis of he Morse Indices of he soluions obained by using he Mounain Pass Theorem
More informationMath 334 Fall 2011 Homework 11 Solutions
Dec. 2, 2 Mah 334 Fall 2 Homework Soluions Basic Problem. Transform he following iniial value problem ino an iniial value problem for a sysem: u + p()u + q() u g(), u() u, u () v. () Soluion. Le v u. Then
More informationOn Multicomponent System Reliability with Microshocks - Microdamages Type of Components Interaction
On Mulicomponen Sysem Reliabiliy wih Microshocks - Microdamages Type of Componens Ineracion Jerzy K. Filus, and Lidia Z. Filus Absrac Consider a wo componen parallel sysem. The defined new sochasic dependences
More information8. Basic RL and RC Circuits
8. Basic L and C Circuis This chaper deals wih he soluions of he responses of L and C circuis The analysis of C and L circuis leads o a linear differenial equaion This chaper covers he following opics
More informationLecture 2-1 Kinematics in One Dimension Displacement, Velocity and Acceleration Everything in the world is moving. Nothing stays still.
Lecure - Kinemaics in One Dimension Displacemen, Velociy and Acceleraion Everyhing in he world is moving. Nohing says sill. Moion occurs a all scales of he universe, saring from he moion of elecrons in
More informationFailure of the work-hamiltonian connection for free energy calculations. Abstract
Failure of he work-hamilonian connecion for free energy calculaions Jose M. G. Vilar 1 and J. Miguel Rubi 1 Compuaional Biology Program, Memorial Sloan-Keering Cancer Cener, 175 York Avenue, New York,
More informationarxiv: v1 [math.ca] 15 Nov 2016
arxiv:6.599v [mah.ca] 5 Nov 26 Counerexamples on Jumarie s hree basic fracional calculus formulae for non-differeniable coninuous funcions Cheng-shi Liu Deparmen of Mahemaics Norheas Peroleum Universiy
More informationPredator - Prey Model Trajectories and the nonlinear conservation law
Predaor - Prey Model Trajecories and he nonlinear conservaion law James K. Peerson Deparmen of Biological Sciences and Deparmen of Mahemaical Sciences Clemson Universiy Ocober 28, 213 Ouline Drawing Trajecories
More informationCash Flow Valuation Mode Lin Discrete Time
IOSR Journal of Mahemaics (IOSR-JM) e-issn: 2278-5728,p-ISSN: 2319-765X, 6, Issue 6 (May. - Jun. 2013), PP 35-41 Cash Flow Valuaion Mode Lin Discree Time Olayiwola. M. A. and Oni, N. O. Deparmen of Mahemaics
More informationNotes on Kalman Filtering
Noes on Kalman Filering Brian Borchers and Rick Aser November 7, Inroducion Daa Assimilaion is he problem of merging model predicions wih acual measuremens of a sysem o produce an opimal esimae of he curren
More informationPENALIZED LEAST SQUARES AND PENALIZED LIKELIHOOD
PENALIZED LEAST SQUARES AND PENALIZED LIKELIHOOD HAN XIAO 1. Penalized Leas Squares Lasso solves he following opimizaion problem, ˆβ lasso = arg max β R p+1 1 N y i β 0 N x ij β j β j (1.1) for some 0.
More informationPhysics 127b: Statistical Mechanics. Fokker-Planck Equation. Time Evolution
Physics 7b: Saisical Mechanics Fokker-Planck Equaion The Langevin equaion approach o he evoluion of he velociy disribuion for he Brownian paricle migh leave you uncomforable. A more formal reamen of his
More informationLecture 20: Riccati Equations and Least Squares Feedback Control
34-5 LINEAR SYSTEMS Lecure : Riccai Equaions and Leas Squares Feedback Conrol 5.6.4 Sae Feedback via Riccai Equaions A recursive approach in generaing he marix-valued funcion W ( ) equaion for i for he
More information5. Stochastic processes (1)
Lec05.pp S-38.45 - Inroducion o Teleraffic Theory Spring 2005 Conens Basic conceps Poisson process 2 Sochasic processes () Consider some quaniy in a eleraffic (or any) sysem I ypically evolves in ime randomly
More informationOn Boundedness of Q-Learning Iterates for Stochastic Shortest Path Problems
MATHEMATICS OF OPERATIONS RESEARCH Vol. 38, No. 2, May 2013, pp. 209 227 ISSN 0364-765X (prin) ISSN 1526-5471 (online) hp://dx.doi.org/10.1287/moor.1120.0562 2013 INFORMS On Boundedness of Q-Learning Ieraes
More informationModule 2 F c i k c s la l w a s o s f dif di fusi s o i n
Module Fick s laws of diffusion Fick s laws of diffusion and hin film soluion Adolf Fick (1855) proposed: d J α d d d J (mole/m s) flu (m /s) diffusion coefficien and (mole/m 3 ) concenraion of ions, aoms
More informationEchocardiography Project and Finite Fourier Series
Echocardiography Projec and Finie Fourier Series 1 U M An echocardiagram is a plo of how a porion of he hear moves as he funcion of ime over he one or more hearbea cycles If he hearbea repeas iself every
More informationACE 562 Fall Lecture 4: Simple Linear Regression Model: Specification and Estimation. by Professor Scott H. Irwin
ACE 56 Fall 005 Lecure 4: Simple Linear Regression Model: Specificaion and Esimaion by Professor Sco H. Irwin Required Reading: Griffihs, Hill and Judge. "Simple Regression: Economic and Saisical Model
More informationUnit Root Time Series. Univariate random walk
Uni Roo ime Series Univariae random walk Consider he regression y y where ~ iid N 0, he leas squares esimae of is: ˆ yy y y yy Now wha if = If y y hen le y 0 =0 so ha y j j If ~ iid N 0, hen y ~ N 0, he
More informationSome Basic Information about M-S-D Systems
Some Basic Informaion abou M-S-D Sysems 1 Inroducion We wan o give some summary of he facs concerning unforced (homogeneous) and forced (non-homogeneous) models for linear oscillaors governed by second-order,
More informationIn this chapter the model of free motion under gravity is extended to objects projected at an angle. When you have completed it, you should
Cambridge Universiy Press 978--36-60033-7 Cambridge Inernaional AS and A Level Mahemaics: Mechanics Coursebook Excerp More Informaion Chaper The moion of projeciles In his chaper he model of free moion
More informationMath Week 14 April 16-20: sections first order systems of linear differential equations; 7.4 mass-spring systems.
Mah 2250-004 Week 4 April 6-20 secions 7.-7.3 firs order sysems of linear differenial equaions; 7.4 mass-spring sysems. Mon Apr 6 7.-7.2 Sysems of differenial equaions (7.), and he vecor Calculus we need
More informationConvergence of the Neumann series in higher norms
Convergence of he Neumann series in higher norms Charles L. Epsein Deparmen of Mahemaics, Universiy of Pennsylvania Version 1.0 Augus 1, 003 Absrac Naural condiions on an operaor A are given so ha he Neumann
More informationOn Measuring Pro-Poor Growth. 1. On Various Ways of Measuring Pro-Poor Growth: A Short Review of the Literature
On Measuring Pro-Poor Growh 1. On Various Ways of Measuring Pro-Poor Growh: A Shor eview of he Lieraure During he pas en years or so here have been various suggesions concerning he way one should check
More informationM-estimation in regression models for censored data
Journal of Saisical Planning and Inference 37 (2007) 3894 3903 www.elsevier.com/locae/jspi M-esimaion in regression models for censored daa Zhezhen Jin Deparmen of Biosaisics, Mailman School of Public
More informationA Uniform Asymptotic Estimate for Discounted Aggregate Claims with Subexponential Tails
A Uniform Asympoic Esimae for Discouned Aggregae Claims wih Subeponenial Tails Xuemiao Hao and Qihe Tang Deparmen of Saisics and Acuarial Science The Universiy of Iowa 241 Schae er Hall, Iowa Ciy, IA 52242,
More informationNotes for Lecture 17-18
U.C. Berkeley CS278: Compuaional Complexiy Handou N7-8 Professor Luca Trevisan April 3-8, 2008 Noes for Lecure 7-8 In hese wo lecures we prove he firs half of he PCP Theorem, he Amplificaion Lemma, up
More informationDifferential Equations
Mah 21 (Fall 29) Differenial Equaions Soluion #3 1. Find he paricular soluion of he following differenial equaion by variaion of parameer (a) y + y = csc (b) 2 y + y y = ln, > Soluion: (a) The corresponding
More informationTHE BERNOULLI NUMBERS. t k. = lim. = lim = 1, d t B 1 = lim. 1+e t te t = lim t 0 (e t 1) 2. = lim = 1 2.
THE BERNOULLI NUMBERS The Bernoulli numbers are defined here by he exponenial generaing funcion ( e The firs one is easy o compue: (2 and (3 B 0 lim 0 e lim, 0 e ( d B lim 0 d e +e e lim 0 (e 2 lim 0 2(e
More informationMatlab and Python programming: how to get started
Malab and Pyhon programming: how o ge sared Equipping readers he skills o wrie programs o explore complex sysems and discover ineresing paerns from big daa is one of he main goals of his book. In his chaper,
More information20. Applications of the Genetic-Drift Model
0. Applicaions of he Geneic-Drif Model 1) Deermining he probabiliy of forming any paricular combinaion of genoypes in he nex generaion: Example: If he parenal allele frequencies are p 0 = 0.35 and q 0
More information1 Review of Zero-Sum Games
COS 5: heoreical Machine Learning Lecurer: Rob Schapire Lecure #23 Scribe: Eugene Brevdo April 30, 2008 Review of Zero-Sum Games Las ime we inroduced a mahemaical model for wo player zero-sum games. Any
More informationKINEMATICS IN ONE DIMENSION
KINEMATICS IN ONE DIMENSION PREVIEW Kinemaics is he sudy of how hings move how far (disance and displacemen), how fas (speed and velociy), and how fas ha how fas changes (acceleraion). We say ha an objec
More informationdy dx = xey (a) y(0) = 2 (b) y(1) = 2.5 SOLUTION: See next page
Assignmen 1 MATH 2270 SOLUTION Please wrie ou complee soluions for each of he following 6 problems (one more will sill be added). You may, of course, consul wih your classmaes, he exbook or oher resources,
More informationTwo Coupled Oscillators / Normal Modes
Lecure 3 Phys 3750 Two Coupled Oscillaors / Normal Modes Overview and Moivaion: Today we ake a small, bu significan, sep owards wave moion. We will no ye observe waves, bu his sep is imporan in is own
More informationMonotonic Solutions of a Class of Quadratic Singular Integral Equations of Volterra type
In. J. Conemp. Mah. Sci., Vol. 2, 27, no. 2, 89-2 Monoonic Soluions of a Class of Quadraic Singular Inegral Equaions of Volerra ype Mahmoud M. El Borai Deparmen of Mahemaics, Faculy of Science, Alexandria
More information18 Biological models with discrete time
8 Biological models wih discree ime The mos imporan applicaions, however, may be pedagogical. The elegan body of mahemaical heory peraining o linear sysems (Fourier analysis, orhogonal funcions, and so
More informationCERTAIN CLASSES OF SOLUTIONS OF LAGERSTROM EQUATIONS
SARAJEVO JOURNAL OF MATHEMATICS Vol.10 (22 (2014, 67 76 DOI: 10.5644/SJM.10.1.09 CERTAIN CLASSES OF SOLUTIONS OF LAGERSTROM EQUATIONS ALMA OMERSPAHIĆ AND VAHIDIN HADŽIABDIĆ Absrac. This paper presens sufficien
More informationSOLUTIONS TO ECE 3084
SOLUTIONS TO ECE 384 PROBLEM 2.. For each sysem below, specify wheher or no i is: (i) memoryless; (ii) causal; (iii) inverible; (iv) linear; (v) ime invarian; Explain your reasoning. If he propery is no
More informationPhysics 235 Chapter 2. Chapter 2 Newtonian Mechanics Single Particle
Chaper 2 Newonian Mechanics Single Paricle In his Chaper we will review wha Newon s laws of mechanics ell us abou he moion of a single paricle. Newon s laws are only valid in suiable reference frames,
More informationCURRENT STATUS DATA WITH COMPETING RISKS: LIMITING DISTRIBUTION OF THE MLE. By Piet Groeneboom, Marloes H. Maathuis. and Jon A.
CURRENT STATUS DATA WITH COMPETING RISKS: LIMITING DISTRIBUTION OF THE MLE By Pie Groeneboom, Marloes H. Maahuis and Jon A. Wellner Delf Universiy of Technology and Vrije Universiei Amserdam, Universiy
More information1 Solutions to selected problems
1 Soluions o seleced problems 1. Le A B R n. Show ha in A in B bu in general bd A bd B. Soluion. Le x in A. Then here is ɛ > 0 such ha B ɛ (x) A B. This shows x in B. If A = [0, 1] and B = [0, 2], hen
More informationLecture 6: Wiener Process
Lecure 6: Wiener Process Eric Vanden-Eijnden Chapers 6, 7 and 8 offer a (very) brief inroducion o sochasic analysis. These lecures are based in par on a book projec wih Weinan E. A sandard reference for
More informationRobotics I. April 11, The kinematics of a 3R spatial robot is specified by the Denavit-Hartenberg parameters in Tab. 1.
Roboics I April 11, 017 Exercise 1 he kinemaics of a 3R spaial robo is specified by he Denavi-Harenberg parameers in ab 1 i α i d i a i θ i 1 π/ L 1 0 1 0 0 L 3 0 0 L 3 3 able 1: able of DH parameers of
More informationCURRENT STATUS DATA WITH COMPETING RISKS: LIMITING DISTRIBUTION OF THE MLE. By Piet Groeneboom, Marloes H. Maathuis. and Jon A.
CURRENT STATUS DATA WITH COMPETING RISKS: LIMITING DISTRIBUTION OF THE MLE By Pie Groeneboom, Marloes H. Maahuis and Jon A. Wellner Delf Universiy of Technology and Vrije Universiei Amserdam, Universiy
More information