Divergence criteria for improved selection rules
|
|
- Mitchell Mathews
- 5 years ago
- Views:
Transcription
1 Divergece criteria for improved selectio rules. Berliet ad I. Vajda N ovem b er 3, 00 bstract t the basis of combiatorial methods i desity estimatio itroduced by Devroye ad Lugosi is the so-called Sche é selectio rule. We show by a examples that this rule based o L errors may ot brig the selectio closer to optimality tha tossig of a coi. s i ay estimatio problem, the choice of a criterio is at the heart of the matter. The optimality of the Sche é estimate is perceived di eretly by di eret divergece criteria. We show that the L oracle iequality satis ed by the Sche é estimate ca be exteded to divergeces. It ca be also exteded to estimates associated with selectio rules based o divergeces. s the L rule, the ew rules are applicable to ay selectio problem i desity estimatio. MS 99 subject classi catio: 6 G 05. Key Words: Desity estimatio. Noparametric estimatio. Selectio of estimates. Iformatio divergece. Optimality. Combiatorial methods. Itroductio ad basic cocepts I the book of Devroye ad Lugosi (00), the authors cosidered the statistical model with R d -valued observatios X ; : : : ; X i.i.d. by a probability desity f o R d ad two estimates f (i) = f (i) (; X ; : : : ; X ); i f; g () of this desity. They were iterested i the problem how to select for each realizatio of X ; : : : ; X the better of these estimates i the sese of L -error. Obviously, the optimal but practically uachievable selectio is f (0) = >< >: f () f () if otherwise: jf () fj < They proposed a practically achievable approximatio to this selectio called Sche é estimate which is selected by by the rule >< f () f if f () ( ) = < f () ( ) ; (3) >: otherwise f () jf () I3M, UMR CNRS 549, Uiversity of Motpellier II, Motpellier Cedex, Frace Istitute of Iformatio Theory ad utomatio, SCR, 0 Prague, Czech Republic fj; ()
2 where is the so-called Sche é set for f () ; f () ad = f () ; f () = x : f () (x) > f () (x) (4) () = X I (X i ) ; B d ; (5) i= is the empirical probability measure o the -algebra of Borel sets B d. Chapter 6 of the cited book cotais a umber of argumets i favour of the Sche é selectio rule (3). However, the ext example demostrates that the favorizatio of the Sche é rule is problematic i some cases. s above, I () deotes the idicator fuctio. Example. Cosider f uiform o the closed iterval [0; ] R ad the correspodig ordered sample X : ; : : : ; X : : For the estimates of f we get so that f () = I(X : x X : + ) ad f () = I(X : x X : ) = (X : ; X : + j; ( ) = 0 f () = X : + X : ad f () = 0 f () ( ) = jx : + X : j > X : exceeds with probability the absolute deviatio ( ) = 0: f () Cosequetly the Sche é rule selects the estimate f () achievig the L -error R jf () fj = ( X : ) whereas the estimate f () achieves the error R jf () fj = X : ; so that is strictly better i the L -sese with the probability Pr(X : + X : < ) = =. The book of Devroye ad Lugosi (00) presets a systematic theory dealig with properties ad applicatios of the Sche é selectio f: This theory is based o Theorem 6. which compares the errors jf (0) fj = mi jf () fj; This fudametal theorem ca be give the form of the iequality jf fj 3 jf (0) fj + 4 f ( ) (6) jf () fj ad jf fj:
3 where is the Sche é set for f () ; f (). This iequality states that the selectio f ca achieve the error level 3 R jf (0) fj up to the uiversal error term appearig o the right. This iequality was applied ot oly i Chapters 7 7 of the Devroye-Lugosi book, but also i subsequet papers, amog them i Berliet, Biau ad Rouviere (005 a,b). The latter papers observed that the L -error criterio R jf gj for the estimates g beig formally probability desities is a special case of the more geeral -divergece criterio D (f; g) de ed for arbitrary probability desities f; g by the formula f D (f; g) = g : (7) g Here (t) is oegative ad covex i the domai t (0; ), strictly covex ad vaishig at the poit t = (for details about formula (7) ad the basic properties of -divergeces used below, see Csiszár (967a) or Liese ad Vajda (97, 006). The L -error is the -divergece for (t) = jt j, called total variatio ad deoted by V (f; g), i. e. V (f; g) = jf gj = sup f g : () B d Other examples are the squared Helliger distace the squared Le Cam distace H (f; g) = p f p g for (t) = p t, (9) LC (f; g) = ad the iformatio divergece I(f; g) = (f g) f + g f l f g for (t) = (t ) (t + ) ; (0) for (t) = t l t: () Natural motivatio for the alterative -divergece error criteria is the eed to work with estimates covereget i the topologies stroger tha that iduced by the total variatio (cf. Csiszár 967b ad Österreicher ad Vajda (003). This paper itroduces a ew motivatio achieved i Example 3 below by extedig the framework of Example through admittig o-uiform desities with uit supports o R: I this exteded settig Example 3 demostrates that for some desities f the alterative -divergece error criteria exhibit with positive probabilities optimality of the estimate g = f () at the same time whe the L -error exhibits the optimality of g = f (). Sice the optimality of the Sche é estimates f is perceived di eretly by di eret - divergece error criteria, it is importat to see whether or how the fudametal Devroye Lugosi iequality (6) ca be exteded from the total variatio criteria V (f; f ) = jf fj ad V (f; f (0) ) = 3 jf (0) fj ()
4 to the more geeral -divergece criteria D (f; f) = f f f This problem is solved i Sectio 3. ad D (f; f (0) ) = f (0) f f (0)! : (3) Sectio 4 itroduces a replacemet of the Sche é L -based selectio rule by a more geeral -divergece selectio rule ad solves a problem parallel to that of Sectio 3, amely whether or how the Devroye Lugosi iequality (6) ca be exteded to these estimates ad to the more geeral -divergece criteria. For the obvious reasos, i this paper the attetio is restricted to the estimates () which are a.s. probability desities themselves. Metric divergece criteria of errors Let us start with the followig basic properties of -divergeces eeded i the sequel: (i) The rage of values is 0 D (f; g) (0) + (0) (4) where (0); (0) are smooth extesios of (t), (t) = t (=t) to the poit t = 0. I (4) D (f; g) = 0 if ad oly if f = g a. s. ad D (f; g) = (0) + (0) if (for ite (0) + (0) if ad oly if) f?g (disjoit supports). (ii) The symmetry D (f; g) = D (g; f) for all f; g holds if ad oly if = fot the adjoit fuctio de ed i (i). (iii) The mootoicity property deals with relatios betwee -divergeces of the distributios () = D (; ) D (f; g) f; () = g; B d ad -divergeces of restrictios of these distributios o sub--algebras S B d of the Borel -algebra B d de ed by formula fs D (; js) = D (f S ; g S ) = g S for S-measurable versios f S ; g S of desities f; g. It states that the orderig D (f; gjs) D (; js) D (; ) D (f; g) (5) holds. If the equality i (5) takes place the we say that S preserves the -divergece D (f; g). It is kow (see e.g. Corollary.9 i Liese ad Vajda (97)) that if a sub- -algebra S is su ciet for the pair ff; gg the the equality takes place i (5), i.e. the 4 g S
5 su ciet S allways preserves the -divergece D (f; g). (iv) Fially, the spectral represetatio says that if a sub--algebra S B d is geerated by a ite or coutable B d -measurable partitio P of R d (spectrum of S, i symbols we write S = S(P)) the D (f; gjs) = X P R g : R f g : (6) Example. Cosider for every B d the partitio P = (; c ) of R d ad the P- geerated (or, more simply, -geerated) algebra S := S (; c ) B d (7) cosistig of the sets R d ; ; c ; ;: The the geeral spectral represetatio (6) implies V (f; gjs ) = X f g = f g : () Bf; c g B From () ad () we see that the fudametal Devroye Lugosi iequality (6) ca be give the form V (f; f) 3V (f; f (0) ) + V (; js ) (9) for the Sche é set of the estimates f () ad f () : If i () is the Sche é set (f; g) of f ad g the the absolute di erece o the right of () ca be replaced by the ordiary di erece. Moreover, it is see from () that the S preserves V (f; g) so that the formula () ca be exteded ad speci ed as follows V (f; gjs ) = V (f; g) = B f g : (0) The followig sectios exted the Devroye Lugosi theorem (6), or equivaletly (9), to the error criteria D(f; g) for probability desities f; g o (R d ; B d ) satisfyig similar metric properties as the total variatio criterio V (f; g) amely the re exivity the symmetry ad the triagle iequality D(f; g) = 0 if ad oly if f = g a. s.; () D(f; g) = D(g; f) for all f; g () D(f; g) D(f; h) + D(h; g) for all f; g; h: (3) We restrict ourselves to the metric divergece criteria de ed as powers D(f; g) = D (f; g) ; > 0 5
6 of -divergeces D (f; g) satisfyig ()-(3). These-divergeces achieve ite uppr bouds (0) + (0) = (0) < (4) (see (ii) for the equality ad Csiszár (967b) for the iteess). To provide a su cietly rich class of such criteria, let us itroduce the class of - divergeces D (f; g) = D (f; g); R: (5) Here the covex fuctios (t) are give i the domai t > 0 by the formula (t) = j j ( ) (t + ) (t = + ) (6) if ( ) 6= 0; ad by the correspodig limits 0 (t) = j t j =; (7) (t) = t l t + (t + ) l () t + otherwise. The subclass of these divergeces for 0 was proposed (with a di eret parametrizatio) by Österreicher ad Vajda (003). The extesio to < 0 was proposed recetly by Vajda (00). It is easy to verify for all f; g the formulas ad D 0 (f; g) = V (f; g) (total variatio, ()); (9) D (f; g) = H (f; g) (Helliger, (9)); (30) D (f; g) = 4 LC (f; g) (Le Cam, (0)) (3) D (f; g) = I (f; (f + g)=) + I(g; (f + g)=): (3) I the ppedix we demostrate that the powers D(f; g) := D (f; g) = maxf;g, R (33) of the divergeces (5) satisfy () (3), i. e. that they are metric divergece criteria. 3 Sche é selectio rule This sectio exteds the fudametal Devroye Lugosi iequality (6) for the Sche é estimates f from the total variatio error criteria () to the more geeral -divergece criteria (3). The ext example provides a motivatio for this extesio. 6
7 Example 3. Mai result of this sectio is the followig theorem. This theorem ad its proof refer to the lower ad upper error bouds L (V ) D (f; g) U (V ) (34) achieved for a give covex by the -divergeces D (f; g) o the class of desities f; g satisyig the total variatio coditio V (f; g) = V; 0 V : By Propositio.7 i Liese ad Vajda (97), the upper boud is for geeral give by the formula U (V ) = V c where c = (0) + (0) (cf. (4)) (35) ad the lower boud L (V ) is covex ad strictly icreasig i the variable V from the miimum L (0) = 0 to the maximum L () = (0) + (0) = c. Hece the strictly icreasig ad cocave iverse fuctio L (D) : [0; c ]! [0; ] (36) allways exists. For such that the powers D (f; g) are metrics o the space of desities f; g (4) implies c = (0) < : (37) Theorem. Let f be a estimated distributio o R d, f (0) ; f () ad f () the estimates cosidered i (), () with the correspodig Sche é set ad f the Sche é estimate resultig from the selectio rule (3). The for every metric divergece criterio D(f; g) = D (f; g) D (f; f) D f (0) ; f + c L D f (0) where L ad c are give by (36) ad (37) ; f = + f ( ) (3) Proof. Cosider the radom variables E ij = I f = f (i) ; f (0) = f (j) where X i;j= E ij = : (39) By the triagle iequality ad symmetry of D(f; g), ad by the de itio of E ii ; D (f ; f) D f (0) ; f + It su ces to prove that for i 6= j D f; f (0) Eij c X i;j= D f; f (0) Eij = D f (0) ; f + D f; f (0) E + D f; f (0) E : (40) L D f (0) ; f = + f ( ) E ij : (4) 7
8 We restrict ourselves to E : For E the proof is similar. By the de itio of E ad (35), (36), D f; f (0) E = D f () ; f () E c V f () ; f () E = c V f () ; f () js E c V f () ; fjs + V f () ; fjs E c V f () ; f + V () ; js + V ( ; js ) E c V f (0) ; f + V ( ; js ) E h c D f (0) ; f = + V ( ; js )i E : L where we bouded the sum of the total variatios i the third lie above by This completes the proof. V f () ; f + V () ; js + V ( ; js ) V f () ; f + V () ; js + V ( ; js ) V f () ; f + V ( ; js ) : The ext corollary reformulates the result of Theorem i a simpler but slightly weaker form. Corollary. For 0 <, uder the assumptios ad otatios of Theorem, D (f; f) c 3L D f (0) ; f = + 4 f ( ) (4) Proof. Clear from (3) by takig ito accout the iequalities D f (0) ; f U V f (0) ; f = c V f (0) ; f h c L D f (0) ; f i = obtaied from (34), (36) ad also the iequality (a) + (b) (a + b) obtaied from Jese s iequality for the cocave fuctio (x) = x. The ext example demostrates that Theorem geeralizes the Devroye ad Lugosi iequality (6). Example 4. Put D(f; g) = D 0 (f; g) = V (f; g)= (cf. (9)). The c 0 = 0 (0) = =; L 0 (V ) = U 0 (V ) = V ; 0 V ad L 0 (D) = D. Hece Theorem implies D 0 (f; f) D 0 f (0) ; f + D 0 f (0) ; f + f ( )
9 or, equivaletly, which coicides with (6) ad (9). V (f; f) 3V f (0) ; f + 4 f ( ) The ext example illustrates cotributios of Theorem ad its Corollary beyod the framework of Devroye ad Lugosi. Example 5. Put D(f; g) = D (f; g) =, i.e. take the Le Cam error criterio LC(f; g)= (cf. (3)). The parts (ii) ad (iii) of Theorem i the ppedix imply that c = =, U (V ) = V=6 ad L (V ) =! + V= + V= = (V=) V = : 4 Therefore L (D) = 4 p D ad for the Sche é selectio f of Devroye ad Lugosi we get from Theorem the relatio D (f; f) D f (0) ; f + 4D f (0) ; f = + f ( ) i. e. LC (f ; f) LC f (0) ; f + r LC f (0) ; f + jf ( )j where is the Sche é set of the iitial estimates f () ad f (). Corollary implies for the same f ad as before the alterative iequality i. e. D (f ; f) LC (f ; f) 3 (0) 4D f ; f + 4 s 3 LC f (0) ; f + 4 f ( ) = f ( ) : We see that the rate of covergece of the Le Cam error LC (f; f) to zero garateed by our theory for the Sche é estimate is strictly below the rate of the Le Cam error LC ; f achieved by the ideal estimate f (0). Oe ca deduce from the kow properties of the lower boud L (V ) ad its iverse L (D) that similar result ca be expected also for other divergece errors D (f; f) with strictly covex everywhere. f (0) 9
10 4 Divergece selectio rule This sectio is a cotiuatio of Sectio 3 where the estimatio errors are still evaluated by metric divergece criteria of the type D(f; g) = D (f; g) ; > 0 but the Sche é selectio (3) of Devroye ad Lugosi (00) is replaced by a more geeral selectio. Oe arrives quite aturally at such geeralizatio if he applies the same criteria also to the de itio of the optimal estimate f (0) ad its practical approximatio f. I other words, the geeralizatio cosists i the replacemet of the L -based de itio () by the divergece based de itio < f (0) = : f () f () if D f () otherwise: ; f < D f () ; f ad the L -based Sche é selectio rule (3) by the divergece selectio rule < f () f if D () ; js < D () ; js = : otherwise: f () This rule uses the empirical distributio de ed by (5), the estimates (i) (B) = f (i) ; B B d ; i f; g E of the probability distributio f; ad the sub--algebra S B d preservig the divergece D () ; (), i. e. satisfyig the equality D () ; () = D () ; () js (43) (44) (cf. (5)): (45) Next follows the mai result of this sectio where f is the estimated distributio ad S is sub--algebra preservig the divergece D(f () ; f () ) of the estimates f () ; f () i (43) ( e.g. the itersectio of all sub--algebras S B d preservig this divergece). Theorem. The estimate f resultig from the metric divergece selectio rule (44) satis es the iequality D (f ; f) 3D f (0) ; f + D (; js ) : (46) Proof. We ca start with the equality (40) valid i the preset situatio as well. It su ces to prove that for i 6= j D f; f (0) Eij D f (0) ; f + D ( ; js ) E ij 0
11 where E ij is de ed by (39) for f; f (0) give by (43), (44). Usig repeatedly the triagle iequality ad relatios (45) ad (5) we obtai D f ; f (0) E = D f (0) ; f () D f () E = D f () ; fjs + D f () ; f () js E ; fjs E D f () ; f + D () ; js + D ( ; js ) E D f () ; f + D () ; js + D ( ; js ) E D f () ; f + D () ; js + D ( ; js ) E D f () ; f + D ( ; js ) E = D f (0) ; f + D ( ; js ) E : I the same maer we obtai D f ; f (0) which completes the proof of (46). E D f (0) ; f + D ( ; js ) E The ext corollary presets a di eret expressio of the error term i (46). Corollary. The estimate f resultig from the selectio rule (44) employig a metric divergece D(f; g) = D (f; g) satis es the iequality D (f; f) 3D f (0) ; f + + c sup f (B) (47) BS where f (0) ; f ad S are the same as i Theorem ad c = (0) < : Proof. By Propositio.7 i Liese ad Vajda (97) ad (), D ( ; js ) c V ( ; js ) ad V ( ; js ) = sup S j() ()j ad the rest follows from Theorem ad (37).. s i the previous sectio, our rst step is to verify that Theorem geeralizes the Devroye Lugosi result (6). Example 6. Puttig D(f; g) = V (f; g) i Theorem ad usig the fact that by (0) the sub--algebra S preserves the total variatio V (f () ; f () ) of the estimates f () ; f () ; we get V (f; f) 3V f (0) ; f + V (; js ) : This coicides with the equivalet form (9) of the Devroye-Lugosi iequality (6). Most importat from the poit of view of applicatios is the complexity of the sub-algebra S B d which appears i the right-had error terms of (46) ad (47). It depeds
12 o the complexity of the used divergece error criterio D(f; g) ad the complexity of the estimates f () ; f (). I the previous example we have see that if D(f; g) is as simple as the total variatio V (f; g), the S is the simple -algebra S geerated by just oe set the Sche é set of the estimates f () ; f () irrespectively of how complex these estimates are. I the followig example we shall see the opposite extreme, amely simple estimates f () ; f () leadig to a simple -algebra S = S B geerated by just oe set B speci ed by these estimates, irrespectively of how complex the divergece criterio D(f; g) is. More precisely, B does ot deped o this criterio at all. Example 7. Let the sample X ; : : : ; X be govered by a bell-shaped desity f o R ad cosider the sample mea ad variace = X X i ad = i= X (X i ) ; i= ad also the followig cetral cover set Let f be estimated by Cauchy type desities ad where b = B = fx : jx j < 3 g: (4) f () (x) = [ + (x ) ] f () b (x) = I(x B ) [ + (x ) ] arctg 3 = arctg 3 : I (49) we used the fact that the coditio I(x B ) cuts away from f () (x) two tail probabilities of the size 3 3 f () dx = [ + x ] = + arctg( 3) = arctg 3 so that the f () -probability of the sample cetral cover set is /b. The likelihood ratio f () =f () is piecewise costat, f () (x) b if x f () (x) = B 0 otherwise; where b is the ormalizig factor used i (49). Therefore the sub--algebra S B = fr; B ; B; c ;g B geerated by the cetral cover set B of (4) is su ciet for the family ff () ; f () g. By what was said i Sectio, this meas that S B preserves for every (49)
13 covex the -divergece D (f () ; f () ). I other words, the sub--algebra S cosidered i Theorem ad Corollary is S B. Hece, by Theorem ad formula (6), for every metric divergece criterio D(f; g) = D (f; g) with > 0 D (f; f) 3D f (0) ; f + 4 X 3 (B) f R 5 B f : (50) B BfB ;B c g By Corollary, simpler but i geeral weaker variat of the result (50) is the iequality D (f; f) 3D f (0) ; f + + (0) f (B ) : (5) B Next follows a theorem which geeralizes ad makes precise the pheomea observed i the last example. Theorem 3. If the metric divergece criterio D(f; g) is a -divergece power with (t) strictly covex i the whole domai t > 0 the a sub--algebra S B d preserves D(f () ; f () ) i the sese D f () ; f () js = D f () ; f () if ad oly if S is su ciet for ff () ; f () g. Proof. Let D(f () ; f () ) = D (f () ; f () ) for some > 0. By the Corollary above, the metricity of D (f; g) implies D (f () ; f () ) (0) <. Hece, by Corollary.9 i Liese ad Vajda (97), the equality D (f () ; f () ) = D (f () ; f () js ) takes place if ad oly if S is su ciet. From this theorem we see that fuctios strictly covex everywhere de e the most complex divergece criteria for which the -algebra S is simple oly if the estimates f () ; f () are simple eough. Example 4 illustrated such situatio. 5 ppedix For practical applicatios of the results of Sectios 3 ad 4 oe eeds cocrete metric divergece criteria D(f; g) = D (f; g) with kow ad simple upper ad lower boud U (V ) ad L (V ) itroduced i (34). For this purpose he ca use the criteria from theclass >< whe < D(f; g) = D (f; g) () for () = maxf; g = >: whe > : itroduced i (5) (). The followig theorem summarizes basic relevat properties of the divergeces D (f; g). For the proof we refer to Vajda (00). (5) 3
14 Theorem. (i) D (f; g) are -divergeces with fuctios (t) strictly covex i the domai t > 0 whe 6= 0. (ii) The lower bouds of the divergeces D (f; g), R uder the costrait V (f; g) = V are give for 0 V by the formulas L (V ) = jj ( ) " + V = + # =! V (53) if ( ) 6= 0 ad otherwise by the correspodig limits L 0 (V ) = V=; L (V ) = + V l + V + V l V : (54) (iii) The upper bouds of the divergeces D (f; g), R uder the costrait V (f; g) = V are U (V ) = c V where c > 0 is cotiuous i the variable R; give by the formula whe < 0 >< jj + c = (0) = l whe = (55) >: whe 0; 6= : (iv) The powers D (f; g) () give i (5) are metrics i the space of probability desities f; g. Remark. Puttig = 0 i (iv) of Theorem oe obtais amog other the triagle iequality p D0 (f; g) p D 0 (f; h) + p D 0 (h; g) for the divergece D 0 (f; g) = V (f; g)= which is weaker tha the classical triagle iequality D 0 (f; g) D 0 (f; h) + D 0 (h; g) (56) obtaied by applyig the L -orm argumet to the total variatio V (f; g). Usig the cotiuity of the divergeces D (f; g) i the variable R we ca deduce from (56) that more sophisticated argumets tha those used to prove Theorem lead to stroger triagle iequalities also for the remaiig divergeces D (f; g), R, i particular for those with close to 0. ckowledgemet. This research was supported by the grats G µcr 0/07/3 ad MŠMT M05. 4
15 Refereces. Berliet, G. Biau ad L. Rouviere (005a): Parameter selectio i modi ed histogram estimates. Statistics 39, Berliet, G. Biau ad L. Rouviere (005b): Optimal L badwith selectio for variable kerel estimates. Statistics ad Probability Letters 74,.6-7. I. Csiszár (967a): Iformatio-type measures of di erece of probability distributios ad idirect observatios. Studia Sci. Math. Hugarica, I. Csiszár (967b): O topological properties of f-divergeces. Studia Sci. Math. Hugarica, L. Devroye ad G. Lugosi (00): Combiatorial Methods i Desity Estimatio. Spriger, Berli. F. Liese ad I. Vajda (97): Covex Statistical Distaces. Teuber, Leipzig. F. Liese ad I. Vajda (006): O divergeces ad iformatios i statistics ad iformatio theory. IEEE Tras. Iform. Theory 5, 0, F. Österreicher ad I. Vajda (003): ew class of metric divergeces o probability spaces ad its statistical applicatios.. Ist. Statist. Math. 55, I. Vajda (00): O metric f-divergeces of probability measures. Kyberetika (submitted). 5
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 informationLecture 3 The Lebesgue Integral
Lecture 3: The Lebesgue Itegral 1 of 14 Course: Theory of Probability I Term: Fall 2013 Istructor: Gorda Zitkovic Lecture 3 The Lebesgue Itegral The costructio of the itegral Uless expressly specified
More informationEconomics 241B Relation to Method of Moments and Maximum Likelihood OLSE as a Maximum Likelihood Estimator
Ecoomics 24B Relatio to Method of Momets ad Maximum Likelihood OLSE as a Maximum Likelihood Estimator Uder Assumptio 5 we have speci ed the distributio of the error, so we ca estimate the model parameters
More informationThe log-behavior of n p(n) and n p(n)/n
Ramauja J. 44 017, 81-99 The log-behavior of p ad p/ William Y.C. Che 1 ad Ke Y. Zheg 1 Ceter for Applied Mathematics Tiaji Uiversity Tiaji 0007, P. R. Chia Ceter for Combiatorics, LPMC Nakai Uivercity
More informationProduct 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 informationLECTURE 11 LINEAR PROCESSES III: ASYMPTOTIC RESULTS
PRIL 7, 9 where LECTURE LINER PROCESSES III: SYMPTOTIC RESULTS (Phillips ad Solo (99) ad Phillips Lecture Notes o Statioary ad Nostatioary Time Series) I this lecture, we discuss the LLN ad CLT for a liear
More informationIt 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 information7.1 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 informationEquivalence Between An Approximate Version Of Brouwer s Fixed Point Theorem And Sperner s Lemma: A Constructive Analysis
Applied Mathematics E-Notes, 11(2011), 238 243 c ISSN 1607-2510 Available free at mirror sites of http://www.math.thu.edu.tw/ame/ Equivalece Betwee A Approximate Versio Of Brouwer s Fixed Poit Theorem
More informationMeasure 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 informationChapter 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 informationECE 901 Lecture 13: Maximum Likelihood Estimation
ECE 90 Lecture 3: Maximum Likelihood Estimatio R. Nowak 5/7/009 The focus of this lecture is to cosider aother approach to learig based o maximum likelihood estimatio. Ulike earlier approaches cosidered
More informationDimension-free PAC-Bayesian bounds for the estimation of the mean of a random vector
Dimesio-free PAC-Bayesia bouds for the estimatio of the mea of a radom vector Olivier Catoi CREST CNRS UMR 9194 Uiversité Paris Saclay olivier.catoi@esae.fr Ilaria Giulii Laboratoire de Probabilités et
More informationOn equivalent strictly G-convex renormings of Banach spaces
Cet. Eur. J. Math. 8(5) 200 87-877 DOI: 0.2478/s533-00-0050-3 Cetral Europea Joural of Mathematics O equivalet strictly G-covex reormigs of Baach spaces Research Article Nataliia V. Boyko Departmet of
More informationAn 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 informationTR/46 OCTOBER THE ZEROS OF PARTIAL SUMS OF A MACLAURIN EXPANSION A. TALBOT
TR/46 OCTOBER 974 THE ZEROS OF PARTIAL SUMS OF A MACLAURIN EXPANSION by A. TALBOT .. Itroductio. A problem i approximatio theory o which I have recetly worked [] required for its solutio a proof that the
More informationDefinition 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 informationMASSACHUSETTS 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 informationREAL 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(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 informationFIXED POINTS OF n-valued MULTIMAPS OF THE CIRCLE
FIXED POINTS OF -VALUED MULTIMAPS OF THE CIRCLE Robert F. Brow Departmet of Mathematics Uiversity of Califoria Los Ageles, CA 90095-1555 e-mail: rfb@math.ucla.edu November 15, 2005 Abstract A multifuctio
More informationEstimation for Complete Data
Estimatio for Complete Data complete data: there is o loss of iformatio durig study. complete idividual complete data= grouped data A complete idividual data is the oe i which the complete iformatio of
More informationElement sampling: Part 2
Chapter 4 Elemet samplig: Part 2 4.1 Itroductio We ow cosider uequal probability samplig desigs which is very popular i practice. I the uequal probability samplig, we ca improve the efficiecy of the resultig
More informationA class of spectral bounds for Max k-cut
A class of spectral bouds for Max k-cut Miguel F. Ajos, José Neto December 07 Abstract Let G be a udirected ad edge-weighted simple graph. I this paper we itroduce a class of bouds for the maximum k-cut
More informationSOME GENERALIZATIONS OF OLIVIER S THEOREM
SOME GENERALIZATIONS OF OLIVIER S THEOREM Alai Faisat, Sait-Étiee, Georges Grekos, Sait-Étiee, Ladislav Mišík Ostrava (Received Jauary 27, 2006) Abstract. Let a be a coverget series of positive real umbers.
More informationChapter 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 informationOn n-dimensional Hilbert transform of weighted distributions
O -dimesioal Hilbert trasform of weighted distributios MARTHA GUMÁN-PARTIDA Departameto de Matemáticas, Uiversidad de Soora, Hermosillo, Soora 83000, México Abstract We de e a family of cougate Poisso
More informationSequences 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 informationECE 901 Lecture 12: Complexity Regularization and the Squared Loss
ECE 90 Lecture : Complexity Regularizatio ad the Squared Loss R. Nowak 5/7/009 I the previous lectures we made use of the Cheroff/Hoeffdig bouds for our aalysis of classifier errors. Hoeffdig s iequality
More informationJournal of Multivariate Analysis. Superefficient estimation of the marginals by exploiting knowledge on the copula
Joural of Multivariate Aalysis 102 (2011) 1315 1319 Cotets lists available at ScieceDirect Joural of Multivariate Aalysis joural homepage: www.elsevier.com/locate/jmva Superefficiet estimatio of the margials
More informationInfinite 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 informationINEQUALITY FOR CONVEX FUNCTIONS. p i. q i
Joural of Iequalities ad Special Fuctios ISSN: 17-4303, URL: http://www.ilirias.com Volume 6 Issue 015, Pages 5-14. SOME INEQUALITIES FOR f-divergences VIA SLATER S INEQUALITY FOR CONVEX FUNCTIONS SILVESTRU
More informationRandom Variables, Sampling and Estimation
Chapter 1 Radom Variables, Samplig ad Estimatio 1.1 Itroductio This chapter will cover the most importat basic statistical theory you eed i order to uderstad the ecoometric material that will be comig
More informationECE 901 Lecture 14: Maximum Likelihood Estimation and Complexity Regularization
ECE 90 Lecture 4: Maximum Likelihood Estimatio ad Complexity Regularizatio R Nowak 5/7/009 Review : Maximum Likelihood Estimatio We have iid observatios draw from a ukow distributio Y i iid p θ, i,, where
More informationMath 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 informationLet 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 informationBasics of Probability Theory (for Theory of Computation courses)
Basics of Probability Theory (for Theory of Computatio courses) Oded Goldreich Departmet of Computer Sciece Weizma Istitute of Sciece Rehovot, Israel. oded.goldreich@weizma.ac.il November 24, 2008 Preface.
More informationON RECURRENCE COEFFICIENTS FOR RAPIDLY DECREASING EXPONENTIAL WEIGHTS
ON RECURRENCE COEFFICIENTS FOR RAPIDLY DECREASING EXPONENTIAL WEIGHTS E. LEVIN, D. S. LUBINSKY Abstract. Let, for example, W (x) = exp exp k x, x [ ] where > 0, k ad exp k = exp (exp ( exp ())) deotes
More informationMarcinkiwiecz-Zygmund Type Inequalities for all Arcs of the Circle
Marcikiwiecz-ygmud Type Iequalities for all Arcs of the Circle C.K. Kobidarajah ad D. S. Lubisky Mathematics Departmet, Easter Uiversity, Chekalady, Sri Laka; Mathematics Departmet, Georgia Istitute of
More informationBoundaries and the James theorem
Boudaries ad the James theorem L. Vesely 1. Itroductio The followig theorem is importat ad well kow. All spaces cosidered here are real ormed or Baach spaces. Give a ormed space X, we deote by B X ad S
More informationMASSACHUSETTS 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 informationEmpirical Processes: Glivenko Cantelli Theorems
Empirical Processes: Gliveko Catelli Theorems Mouliath Baerjee Jue 6, 200 Gliveko Catelli classes of fuctios The reader is referred to Chapter.6 of Weller s Torgo otes, Chapter??? of VDVW ad Chapter 8.3
More informationApproximation by Superpositions of a Sigmoidal Function
Zeitschrift für Aalysis ud ihre Aweduge Joural for Aalysis ad its Applicatios Volume 22 (2003, No. 2, 463 470 Approximatio by Superpositios of a Sigmoidal Fuctio G. Lewicki ad G. Mario Abstract. We geeralize
More informationChapter 5. Inequalities. 5.1 The Markov and Chebyshev inequalities
Chapter 5 Iequalities 5.1 The Markov ad Chebyshev iequalities As you have probably see o today s frot page: every perso i the upper teth percetile ears at least 1 times more tha the average salary. I other
More informationLecture 3 : Random variables and their distributions
Lecture 3 : Radom variables ad their distributios 3.1 Radom variables Let (Ω, F) ad (S, S) be two measurable spaces. A map X : Ω S is measurable or a radom variable (deoted r.v.) if X 1 (A) {ω : X(ω) A}
More informationIntegrable 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 informationSingular Continuous Measures by Michael Pejic 5/14/10
Sigular Cotiuous Measures by Michael Peic 5/4/0 Prelimiaries Give a set X, a σ-algebra o X is a collectio of subsets of X that cotais X ad ad is closed uder complemetatio ad coutable uios hece, coutable
More information4.1 Data processing inequality
ECE598: Iformatio-theoretic methods i high-dimesioal statistics Sprig 206 Lecture 4: Total variatio/iequalities betwee f-divergeces Lecturer: Yihog Wu Scribe: Matthew Tsao, Feb 8, 206 [Ed. Mar 22] Recall
More informationSelf-normalized deviation inequalities with application to t-statistic
Self-ormalized deviatio iequalities with applicatio to t-statistic Xiequa Fa Ceter for Applied Mathematics, Tiaji Uiversity, 30007 Tiaji, Chia Abstract Let ξ i i 1 be a sequece of idepedet ad symmetric
More informationAn alternative proof of a theorem of Aldous. concerning convergence in distribution for martingales.
A alterative proof of a theorem of Aldous cocerig covergece i distributio for martigales. Maurizio Pratelli Dipartimeto di Matematica, Uiversità di Pisa. Via Buoarroti 2. I-56127 Pisa, Italy e-mail: pratelli@dm.uipi.it
More informationSOME 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 informationMAS111 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 informationA survey on penalized empirical risk minimization Sara A. van de Geer
A survey o pealized empirical risk miimizatio Sara A. va de Geer We address the questio how to choose the pealty i empirical risk miimizatio. Roughly speakig, this pealty should be a good boud for the
More informationIntroduction 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 informationDANIELL AND RIEMANN INTEGRABILITY
DANIELL AND RIEMANN INTEGRABILITY ILEANA BUCUR We itroduce the otio of Riema itegrable fuctio with respect to a Daiell itegral ad prove the approximatio theorem of such fuctios by a mootoe sequece of Jorda
More informationLecture 2: Monte Carlo Simulation
STAT/Q SCI 43: Itroductio to Resamplig ethods Sprig 27 Istructor: Ye-Chi Che Lecture 2: ote Carlo Simulatio 2 ote Carlo Itegratio Assume we wat to evaluate the followig itegratio: e x3 dx What ca we do?
More informationCHAPTER 5 SOME MINIMAX AND SADDLE POINT THEOREMS
CHAPTR 5 SOM MINIMA AND SADDL POINT THORMS 5. INTRODUCTION Fied poit theorems provide importat tools i game theory which are used to prove the equilibrium ad eistece theorems. For istace, the fied poit
More informationResampling Methods. X (1/2), i.e., Pr (X i m) = 1/2. We order the data: X (1) X (2) X (n). Define the sample median: ( n.
Jauary 1, 2019 Resamplig Methods Motivatio We have so may estimators with the property θ θ d N 0, σ 2 We ca also write θ a N θ, σ 2 /, where a meas approximately distributed as Oce we have a cosistet estimator
More informationINEQUALITIES BJORN POONEN
INEQUALITIES BJORN POONEN 1 The AM-GM iequality The most basic arithmetic mea-geometric mea (AM-GM) iequality states simply that if x ad y are oegative real umbers, the (x + y)/2 xy, with equality if ad
More informationDiagonal approximations by martingales
Alea 7, 257 276 200 Diagoal approximatios by martigales Jaa Klicarová ad Dalibor Volý Faculty of Ecoomics, Uiversity of South Bohemia, Studetsa 3, 370 05, Cese Budejovice, Czech Republic E-mail address:
More informationLecture Stat Maximum Likelihood Estimation
Lecture Stat 461-561 Maximum Likelihood Estimatio A.D. Jauary 2008 A.D. () Jauary 2008 1 / 63 Maximum Likelihood Estimatio Ivariace Cosistecy E ciecy Nuisace Parameters A.D. () Jauary 2008 2 / 63 Parametric
More informationREGULARIZATION OF CERTAIN DIVERGENT SERIES OF POLYNOMIALS
REGULARIZATION OF CERTAIN DIVERGENT SERIES OF POLYNOMIALS LIVIU I. NICOLAESCU ABSTRACT. We ivestigate the geeralized covergece ad sums of series of the form P at P (x, where P R[x], a R,, ad T : R[x] R[x]
More informationTheorem 3. A subset S of a topological space X is compact if and only if every open cover of S by open sets in X has a finite subcover.
Compactess Defiitio 1. A cover or a coverig of a topological space X is a family C of subsets of X whose uio is X. A subcover of a cover C is a subfamily of C which is a cover of X. A ope cover of X is
More information1 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 informationSieve Estimators: Consistency and Rates of Convergence
EECS 598: Statistical Learig Theory, Witer 2014 Topic 6 Sieve Estimators: Cosistecy ad Rates of Covergece Lecturer: Clayto Scott Scribe: Julia Katz-Samuels, Brado Oselio, Pi-Yu Che Disclaimer: These otes
More informationAvailable online at J. Math. Comput. Sci. 2 (2012), No. 3, ISSN:
Available olie at http://scik.org J. Math. Comput. Sci. 2 (202, No. 3, 656-672 ISSN: 927-5307 ON PARAMETER DEPENDENT REFINEMENT OF DISCRETE JENSEN S INEQUALITY FOR OPERATOR CONVEX FUNCTIONS L. HORVÁTH,
More information6.3 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 informationf 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 informationLecture 13: Maximum Likelihood Estimation
ECE90 Sprig 007 Statistical Learig Theory Istructor: R. Nowak Lecture 3: Maximum Likelihood Estimatio Summary of Lecture I the last lecture we derived a risk (MSE) boud for regressio problems; i.e., select
More informationOptimally Sparse SVMs
A. Proof of Lemma 3. We here prove a lower boud o the umber of support vectors to achieve geeralizatio bouds of the form which we cosider. Importatly, this result holds ot oly for liear classifiers, but
More informationWeek 10. f2 j=2 2 j k ; j; k 2 Zg is an orthonormal basis for L 2 (R). This function is called mother wavelet, which can be often constructed
Wee 0 A Itroductio to Wavelet regressio. De itio: Wavelet is a fuctio such that f j= j ; j; Zg is a orthoormal basis for L (R). This fuctio is called mother wavelet, which ca be ofte costructed from father
More informationSOME SEQUENCE SPACES DEFINED BY ORLICZ FUNCTIONS
ARCHIVU ATHEATICU BRNO Tomus 40 2004, 33 40 SOE SEQUENCE SPACES DEFINED BY ORLICZ FUNCTIONS E. SAVAŞ AND R. SAVAŞ Abstract. I this paper we itroduce a ew cocept of λ-strog covergece with respect to a Orlicz
More informationPETER HARREMOËS AND CHRISTOPHE VIGNAT
AN ENTROPY POWER INEQUALITY FOR THE BINOMIAL FAMILY PETER HARREMOËS AND CHRISTOPHE VIGNAT Abstract. I this paper, we prove that the classical Etropy Power Iequality, as derived i the cotiuous case, ca
More informationThe Boolean Ring of Intervals
MATH 532 Lebesgue Measure Dr. Neal, WKU We ow shall apply the results obtaied about outer measure to the legth measure o the real lie. Throughout, our space X will be the set of real umbers R. Whe ecessary,
More informationAlgorithms for Clustering
CR2: Statistical Learig & Applicatios Algorithms for Clusterig Lecturer: J. Salmo Scribe: A. Alcolei Settig: give a data set X R p where is the umber of observatio ad p is the umber of features, we wat
More informationRademacher Complexity
EECS 598: Statistical Learig Theory, Witer 204 Topic 0 Rademacher Complexity Lecturer: Clayto Scott Scribe: Ya Deg, Kevi Moo Disclaimer: These otes have ot bee subjected to the usual scrutiy reserved for
More informationInformation Theory Tutorial Communication over Channels with memory. Chi Zhang Department of Electrical Engineering University of Notre Dame
Iformatio Theory Tutorial Commuicatio over Chaels with memory Chi Zhag Departmet of Electrical Egieerig Uiversity of Notre Dame Abstract A geeral capacity formula C = sup I(; Y ), which is correct for
More informationON POINTWISE BINOMIAL APPROXIMATION
Iteratioal Joural of Pure ad Applied Mathematics Volume 71 No. 1 2011, 57-66 ON POINTWISE BINOMIAL APPROXIMATION BY w-functions K. Teerapabolar 1, P. Wogkasem 2 Departmet of Mathematics Faculty of Sciece
More informationBeurling 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 informationMachine Learning Brett Bernstein
Machie Learig Brett Berstei Week 2 Lecture: Cocept Check Exercises Starred problems are optioal. Excess Risk Decompositio 1. Let X = Y = {1, 2,..., 10}, A = {1,..., 10, 11} ad suppose the data distributio
More informationChapter 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 informationFUNDAMENTALS OF REAL ANALYSIS by
FUNDAMENTALS OF REAL ANALYSIS by Doğa Çömez Backgroud: All of Math 450/1 material. Namely: basic set theory, relatios ad PMI, structure of N, Z, Q ad R, basic properties of (cotiuous ad differetiable)
More informationBETWEEN QUASICONVEX AND CONVEX SET-VALUED MAPPINGS. 1. Introduction. Throughout the paper we denote by X a linear space and by Y a topological linear
BETWEEN QUASICONVEX AND CONVEX SET-VALUED MAPPINGS Abstract. The aim of this paper is to give sufficiet coditios for a quasicovex setvalued mappig to be covex. I particular, we recover several kow characterizatios
More informationA RANK STATISTIC FOR NON-PARAMETRIC K-SAMPLE AND CHANGE POINT PROBLEMS
J. Japa Statist. Soc. Vol. 41 No. 1 2011 67 73 A RANK STATISTIC FOR NON-PARAMETRIC K-SAMPLE AND CHANGE POINT PROBLEMS Yoichi Nishiyama* We cosider k-sample ad chage poit problems for idepedet data i a
More informationLecture 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 informationPrecise Rates in Complete Moment Convergence for Negatively Associated Sequences
Commuicatios of the Korea Statistical Society 29, Vol. 16, No. 5, 841 849 Precise Rates i Complete Momet Covergece for Negatively Associated Sequeces Dae-Hee Ryu 1,a a Departmet of Computer Sciece, ChugWoo
More informationw (1) ˆx w (1) x (1) /ρ and w (2) ˆx w (2) x (2) /ρ.
2 5. Weighted umber of late jobs 5.1. Release dates ad due dates: maximimizig the weight of o-time jobs Oce we add release dates, miimizig the umber of late jobs becomes a sigificatly harder problem. For
More informationOn Random Line Segments in the Unit Square
O Radom Lie Segmets i the Uit Square Thomas A. Courtade Departmet of Electrical Egieerig Uiversity of Califoria Los Ageles, Califoria 90095 Email: tacourta@ee.ucla.edu I. INTRODUCTION Let Q = [0, 1] [0,
More informationKernel density estimator
Jauary, 07 NONPARAMETRIC ERNEL DENSITY ESTIMATION I this lecture, we discuss kerel estimatio of probability desity fuctios PDF Noparametric desity estimatio is oe of the cetral problems i statistics I
More informationLecture 2. The Lovász Local Lemma
Staford Uiversity Sprig 208 Math 233A: No-costructive methods i combiatorics Istructor: Ja Vodrák Lecture date: Jauary 0, 208 Origial scribe: Apoorva Khare Lecture 2. The Lovász Local Lemma 2. Itroductio
More informationRegression with quadratic loss
Regressio with quadratic loss Maxim Ragisky October 13, 2015 Regressio with quadratic loss is aother basic problem studied i statistical learig theory. We have a radom couple Z = X,Y, where, as before,
More informationREGRESSION WITH QUADRATIC LOSS
REGRESSION WITH QUADRATIC LOSS MAXIM RAGINSKY Regressio with quadratic loss is aother basic problem studied i statistical learig theory. We have a radom couple Z = X, Y ), where, as before, X is a R d
More informationFall 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 informationIf 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 information7 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 informationSTAT Homework 1 - Solutions
STAT-36700 Homework 1 - Solutios Fall 018 September 11, 018 This cotais solutios for Homework 1. Please ote that we have icluded several additioal commets ad approaches to the problems to give you better
More informationEntropy and Ergodic Theory Lecture 5: Joint typicality and conditional AEP
Etropy ad Ergodic Theory Lecture 5: Joit typicality ad coditioal AEP 1 Notatio: from RVs back to distributios Let (Ω, F, P) be a probability space, ad let X ad Y be A- ad B-valued discrete RVs, respectively.
More informationA Complement to Le Cam s Theorem
A Complemet to Le Cam s Theorem Mark G. Low 1 ad Harriso H. hou Uiversity of Pesylvaia ad Yale Uiversity Abstract This paper examies asymptotic equivalece i the sese of Le Cam betwee desity estimatio experimets
More information1 of 7 7/16/2009 6:06 AM Virtual Laboratories > 6. Radom Samples > 1 2 3 4 5 6 7 6. Order Statistics Defiitios Suppose agai that we have a basic radom experimet, ad that X is a real-valued radom variable
More information