HIGHER THAN SECOND-ORDER APPROXIMATIONS VIA TWO-STAGE SAMPLING

Size: px
Start display at page:

Download "HIGHER THAN SECOND-ORDER APPROXIMATIONS VIA TWO-STAGE SAMPLING"

Transcription

1 Sankhyā : The Indian Journal of Statistics 1999, Volume 61, Series A, Pt. 2, pp HIGHER THAN SECOND-ORDER APPROXIMATIONS VIA TWO-STAGE SAMPING By NITIS MUKHOPADHYAY University of Connecticut, Storrs SUMMARY. We consider the classical fixed-width (= 2d) confidence interval estimation problem for the mean µ of a normal population whose variance σ 2 is unknown, but it is assumed that σ > σ where σ (> 0) is known. Under these circumstances, the seminal two-stage procedure of Stein (1945, 1949) has been recently modified by Mukhopadhyay and Duggan (1997), and that modified methodology was shown to enjoy asymptotic second-order characteristics, similar to those found in Woodroofe (1977) and Ghosh and Mukhopadhyay (1981) in the case of the purely sequential estimation strategies, that is, expanding E(N) and the coverage probability respectively up to the orders o(1) and o(d 2 ) as d 0. In Theorem 1.1, we first obtain expansions of both lower and upper bounds of E(N) up to the order O(d 6 ). In Theorem 1.2, we then provide expansions of the lower and upper bounds for the coverage probability associated with the two-stage procedure of Mukhopadhyay and Duggan (1997) up to the order o(d 4 ), whereas Theorem 1.3 further sharpens this order of approximation to O(d 6 ). These results amount to what may be referred as the third-order approximations and beyond via double sampling. Such results are not available in the case of any existing purely sequential and other multistage estimation strategies. 1. Introduction We consider a sequence of independent and identically distributed random variables X 1, X 2,... from a N(µ, σ 2 ) population where µ R and σɛr + are assumed unknown parameters. Given two preassigned numbers d(> 0) and 0 < α < 1, let us address the classical problem of constructing a confidence interval I for µ such that P {µ I} 1 α for all µ and σ 2, and that the length of I is fixed at 2d. Based on X 1,..., X n where n is predetermined, this problem cannot be solved (Dantzig (1940)). However, Stein (1945, 1949) came up with an ingenious technique involving a two-stage sampling strategy in order to tackle this problem. Paper received. April AMS (1991) subject classification. Primary 6212; secondary Key words and phrases. Fixed-width interval, consistency, third-order expansions, higher-order expansions.

2 third-order approximations and beyond 255 One starts with X 1,..., X m where m( 2) is a fixed initial sample size, and obtains X m = m 1 Σ m i=1 X i, Sm 2 = (m 1) 1 Σ m i=1 (X i X m ) 2. et b m be the upper 50α% point of the Student s t distribution with (m 1) degrees of freedom. Define N = max {m, [ b 2 msm/d 2 2] } (1.1) where [u] = largest integer < u. If N = m, no sampling is needed in the second stage. If N > m, then one samples the difference (N m) in the second stage. Now, based on the combined observations X 1,..., X N we propose the fixed-width confidence interval I N = [ X N ± d ] for µ. Under this scheme, Stein (1945) showed that P {µ I N } 1 α for all fixed µ, σ 2, d and α. This property is known as consistency. If one, however, pretends that σ is known, then for fixed n, P {µ I n } = 2Φ(n 1 2 d/σ) 1, which would be at least (1 α) if n is the smallest integer a 2 σ 2 /d 2 = C, say, where a is the upper 50α% point of the standard normal distribution. Ghosh and Mukhopadhyay (1981) showed that the modified twostage procedure of Mukhopadhyay (1980), where m is allowed to go to infinity at a certain rate as d 0, was first-order efficient, but not second-order efficient, in the asymptotic case. In their terminology, the purely sequential procedure of Ray (1957) and Chow and Robbins (1965) was asymptotically second-order efficient (Woodroofe (1977)), but the consistency property was lost. Even though σ is assumed unknown, throughout this paper, we suppose that σ > σ where σ (> 0) is known. In this situation, Mukhopadhyay and Duggan (1997) appropriately modified the original two-stage procedure (1.1) and this modified methodology was shown to enjoy the consistency property together with the asymptotic second-order characteristics, that is expanding E(N) and the coverage probability respectively up to o(1) and o(d 2 ), which were similar to those enjoyed by the fully sequential estimation strategies under the frequentist paradigm. For a Bayesian overview, one may look at Cohen and Sackrowitz (1984). For small and moderate values of C, the modified methodology worked exceptionally well when, for all practical purposes, σ is not within twenty percent of the magnitude of σ. In this paper, we set out to derive the third- and higher order approximations, that is up to o(d 4 ) and O(d 6 ), for the average sample size as well as the coverage probability associated with the double sampling technique of Mukhopadhyay and Duggan (1997). Since σ > σ where σ (> 0) is known, let us define m = max {m 0, [ a 2 σ/d 2 2] } (1.2) where m 0 ( 2) is a fixed integer. et X 1,..., X m be the pilot observations which provide X m and Sm. 2 Then, let N = max {m, [ b 2 msm/d 2 2] } (1.3)

3 256 nitis mukhopadhyay and we implement this two-stage procedure as we did in the case of (1.1). Finally, based on X 1,..., X N, we consider the fixed-width confidence interval I N = [ X N ± d ] for µ. This methodology was proposed by Mukhopadhyay and Duggan (1997). Obviously, the random variables I(N = n) and X n are independent for all n m, where I( ) stands for the indicator function of ( ), and hence [ ( ) ] P (µɛi N ) = E 2Φ N 1 2 d/σ 1... (1.4) where = E [g (N/C)] ( ) g(x) = 2Φ ax 1 2 1, x > (1.5) The main results are now summarized as follows, while their proofs are deferred to Section 3. We write g (k) (x) to denote d k g(x)/dx k for k = 1,..., 6. Theorem 1.1. For the two-stage procedure (1.2) - (1.3), we have as d 0 : η 0 +η 1 C 1 +η 2 C 2 +O(C 3 ) E(N C) (1+η 0 )+η 1 C 1 +η 2 C 2 +O(C 3 ) where η i 1 = a i (σ 2 /σ 2 )i, i = 1, 2, 3, with a 1, a 2, a 3 defined in (2.3). A property such as η 0 +o(1) E(N C) (1+η 0 )+o(1) would be referred to as second-order approximation or second-order efficiency according to Woodroofe (1977) and Ghosh and Mukhopadhyay (1981), respectively. This result was first pointed out in Mukhopadhyay and Duggan (1997) for the particular procedure on hand. One may also look at Ghosh et al. (1997) for an overview. From Theorem 1.1, we can immediately claim that η 0 + η 1 C 1 + o(d 2 ) E(N C) (1 + η 0 ) + η 1 C 1 + o(d 2 ) which can then quite legitimately be viewed as thirdorder approximation. Theorem 1.1 actually provides significantly higher than third-order approximation for the ASN function. From the proof of this result given in Section 3, it becomes clear that bounds for E(N C) can be derived literally up to any desired order of approximation. For more brevity, we continue up to a remainder term which is O(C 3 ). Theorem 1.2. For the two-stage procedure (1.2) - (1.3), we have as d 0 : (1 α) + C 1 { η 0 g (1) (1) + ( σ 2 σ 2 1) g (2) (1) g (3) (1) } + C 2 { η 1 g (1) (1) B 2g (2) (1) B 3g (3) (1) σ4 σ 4 g(4) (1) } + o(d 4 ) P {µɛi N } (1 α) + C 1 { (η 0 + 1) g (1) (1) + ( σ 2 σ 2 1) g (2) (1) + g (3) (1) } +C 2 { η 1 g (1) (1) B 1g (2) (1) B 4g (3) (1) σ4 σ 4 g(4) (1) } + o(d 4 ) with η 0, η 1 defined in Theorem 1.1, B 1, B 2 defined in emma 2.3, and B 3, B 4 defined in emma 2.4.

4 third-order approximations and beyond 257 The bounds for the coverage probability given in Theorem 1.2 can be referred to as expansions up to third-order approximation. Theorem 1.3. For the two-stage procedure (1.2) - (1.3), we have as d 0 : (1 α) + A 1 C 1 + A 2 C 2 + O(d 6 ) P {µ I N } (1 α) + A 1U C 1 + A 2U C 2 + O(d 6 ) where A 1, A 2 are respectively the same coefficients of C 1, C 2 in the lower bound in Theorem 1.2, and A 1U, A 2U are respectively the same coefficients of C 1, C 2 in the upper bound in Theorem 1.2. The major difference between Theorems 1.2 and 1.3 amounts to significantly sharpening the rate of approximation by replacing the o(d 4 ) term in Theorem 1.2 with the O(d 6 ) term in Theorem 1.3. From Section 3, it will be clear that in order to verify Theorem 1.3 we need two additional emmas beyond what we need in providing a proof of Theorem 1.2. The types of higher order approximations developed in this paper are not available for any problem in the sequential estimation literature. Also, these investigations may provide the crucial impetus in the future to explore possibilities of achieving higher than second-order approximations via purely sequential and other sampling strategies. 2. Auxiliary emmas and Proofs Recall that b m is the upper 50α% point of the Student s t distribution with (m 1) degrees of freedom. From Johnson and Kotz (1970, page 102), one finds the following Cornish-Fisher expansion of b m in terms of a: b m = a a(a2 + 1)(m 1) a(5a4 + 16a 2 + 3)(m 1) a(3a6 + 19a a 2 15)(m 1) 3 + O(m 4 ).... (2.1) From (2.1), the following expansions can be immediately verified: b 2 ma 2 = 1 + a 1 m 1 + a 2 m 2 + a 3 m 3 + O(m 4 ), b 4 ma 4 = 1 + 2a 1 m 1 + (a a 2 )m 2 + 2(a 3 + a 1 a 2 )m 3 + O(m 4 ), b 6 ma 6 = 1 + 3a 1 m 1 + 3(a a 2 )m 2 + (3a 3 + 6a 1 a 2 + a 3 1)m 3 + O(m 4 ),

5 258 nitis mukhopadhyay b 8 ma 8 = 1 + 4a 1 m 1 + 2(3a a 2 )m 2 + 4(a 3 + 3a 1 a 2 + a 3 1)m 3 + O(m 4 ), b 10 m a 10 = 1 + 5a 1 m 1 + 5(2a a 2 )m 2 + 5(a 3 + 4a 1 a 2 + 2a 3 1)m 3 + O(m 4 ), b 12 m a 12 = 1 + 6a 1 m 1 + 3(5a a 2 )m 2 + 2(3a a 1 a a 3 1)m 3 where +O(m 4 ),... (2.2) a 1 = 1 2 (a2 + 1), a 2 = 1 24 (4a4 + 23a ), a 3 = 1 48 (2a6 + 26a a ).... (2.3) The first two lemmas are included here for completeness. These were proved in Mukhopadhyay and Duggan (1997). emma 2.1. For the stopping variable N from (1.3) we have as d 0 : P (N = m) = O (h 1 2 (m 1)) where h = (σ 2 /σ2 ) exp{1 (σ 2 /σ2 )} is a positive proper fraction. emma 2.2. For the stopping variable N from (1.3) we have: (i) C 1 2 (N C) N ( 0, 2σ 2 σ 2 ) as d 0; (ii) C 1 (N C) 2 is uniformly integrable for 0 < d < d 0, with sufficiently small d 0. We now set out to obtain both the lower and upper bounds for some of the positive moments of C 1 2 (N C) in the forms of expansions up to the appropriate order, to be made precise in the form of several lemmas. In the sequel, however, we repeatedly use expansions of some of the positive moments of Sm 2 which are given below. E ( ) Sm 4 = σ { m 1 + 2m 2 + 2m 3 + O(m 4 ) }, E ( ) Sm 6 E ( ) Sm 8 E ( ) Sm 10 E ( ) Sm 12 et us define and one notes that = σ 6 { 1 + 6m m m 3 + O(m 4 ) }, = σ 8 { m m m 3 + O(m 4 ) }, = σ 10 { m m m 3 + O(m 4 ) }, = σ 12 { m m m 3 + O(m 4 ) }.... (2.4) T = max { m, b 2 ms 2 md 2},... (2.5) T N T (2.6)

6 third-order approximations and beyond 259 Also, one has N C T C (2.7) From the proofs given in Mukhopadhyay and Duggan (1997), it is clear that emmas hold when N is replaced with T. Observe also from emma 2.1 that m s P (T = m) converges to zero at an arbitrarily fast rate, whatever be s(> 0) fixed. In the proofs of our emmas that follow, we combine terms such as m s P (T = m) for different values of s, and replace these with one generic expression, namely, O(m r ) or O(C r ) where r(> 0) can be chosen sufficiently large. Another point is to be noted here. A proof of Theorem 1.1 is given in Section 3, but it does not exploit any of the following lemmas, directly or otherwise, and hence in the proofs of such lemmas we use Theorem 1.1 whenever needed as N is replaced with T, that is, a result such as the following: E(T C) = η 0 + η 1 C 1 + η 2 C 2 + O(C 3 ).... (2.8) After looking at the expression in (3.7), it will become clear that in the proof of Theorem 1.2, we would need expansions or bounds for (i) E(N C) up to o(c 1 ), (ii) E{C 1 (N C) 2 } up to o(c 1 ), (iii) E{C 3 (N C) 3 } up to o(c 2 ), and (iv) E{C 2 (N C) 4 } up to o(1). The Theorem 1.1 and emmas provide results which are considerably stronger than the requirements in (i) - (iv) respectively. Next, from (3.8) it will also be clear that in the proof of Theorem 1.3, we would need expansions or bounds for (i) E(N C) up to O(C 2 ), (ii) E{C 1 (N C) 2 } up to O(C 2 ), (iii) E{C 3 (N C) 3 } up to O(C 3 ), (iv) E{C 2 (N C) 4 } up to O(C 1 ), (v) E{C 5 (N C) 5 } up to O(C 3 ), and (vi) E{C 3 (N C) 6 } up to O(1). The Theorem 1.1 provides a result which is stronger than what is needed in (i). The emmas respectively provide exactly what is needed in (ii) - (vi). emma 2.3. For the stopping variable N from (1.3), we have as d 0 : 2σ 2 σ B 1C 1 + O(C 2 ) E { C 1 (N C) 2} 2σ 2 σ B 2C 1 + O(C 2 ) where B 1 = (a a 1 + 2)σ 4 σ 4, B 2 = B 1 + 2η 0 + 1, with η 0 and a 1 respectively defined in Theorem 1.1 and (2.3). Proof. First, we will show that E { C 1 (T C) 2} = 2σ 2 σ 2 + B 1C 1 + O(C 2 ).... (2.9) From (2.5), we can write b 2 ms 2 md 2 T mi(t = m) + b 2 ms 2 md 2,... (2.10)

7 260 nitis mukhopadhyay which implies that b 4 ms 4 md 4 T 2 3m 2 I(T = m) + b 4 ms 4 md (2.11) Now, we utilize the expansions of b 4 m, b 2 m and E(S 4 m) from (2.2), (2.4) and the inequalities from (2.10) - (2.11) to write E { (T C) 2 /C } C 1 [ b 4 md 4 E(S 4 m) + C 2 2mCP (T = m) 2Cb 2 mσ 2 d 2] = C [{ 1 + 2a 1 m 1 + ( a a 2 ) m 2 + O(m 3 ) } {1 + 2m 1 + 2m 2 + O(m 3 )} + 1 2{1 + a 1 m 1 +a 2 m 2 + O(m 3 )} ] + O(m r ) = C [ 2m 1 + (2 + 4a 1 + a 2 1)m 2 + O(m 3 ) ] + O(m r ) Similarly, we obtain = 2σ 2 σ 2 + B 1C 1 + O(C 2 ).... (2.12) E { (T C) 2 /C } C 1 [ b 4 md 4 E(S 4 m) + 3m 2 P (T = m) + C 2 2Cb 2 mσ 2 d 2], and this can be simplified along the lines of (2.12). Thus, (2.9) follows. Next, using (2.6), one obtains E { (T C) 2 /C } 2 E { (N C) 2 /C } E { (T C) 2 /C } + 2E(T/C) + C 1,... (2.13) and then the result follows by combining (2.8), (2.9) and (2.13). emma 2.4. For the stopping variable N from (1.3), we have as d 0 : 6C 1 + B 3 C 2 + O(C 3 ) E { C 3 (N C) 3} 6C 1 + B 4 C 2 + O(C 3 ) where B 3 = q 6η 0 3, B 4 = q + 6σ 2 σ 2 + 6η with q = (8 + 6a 1 )σ 4 σ 4 and η 0, a 1 respectively defined in Theorem 1.1 and (2.3). Proof. First, we will show that From (2.5), observe that E { C 3 (T C) 3} = qc 2 + O(C 3 ).... (2.14) b 6 ms 6 md 6 T 3 7m 3 I(T = m) + b 6 ms 6 md 6,... (2.15)

8 third-order approximations and beyond 261 and combine this with (2.10)- (2.11) and utilize (2.2), (2.4) to write E { C 3 (T C) 3} C [ 3 b 6 md 6 E ( { Sm) 6 3C 3m 2 P (T = m) + b 4 md 4 E ( )} Sm 4 +3C 2 b 2 mσ 2 d 2 C 3] = { 1 + 3a 1 m ( a a 2 ) m 2 + O(m 3 ) } { 1 + 6m m 2 +O(m 3 ) } 3 { 1 + 2a 1 m 1 + ( a a 2 ) m 2 + O(m 3 ) } {1 +2m 1 + 2m 2 + O(m 3 ) } + 3 { 1 + a 1 m 1 + a 2 m 2 +O(m 3 ) } 1 + O(m r ) = { 1 + 3(2 + a 1 )m 1 + ( a 1 + 3a a 2 )m 2 + O(m 3 ) } 3 { 1 + 2(1 + a 1 )m 1 + (2 + 4a 1 + a a 2 )m 2 +O(m 3 ) } + 3 { 1 + a 1 m 1 + a 2 m 2 + O(m 3 ) } 1 + O(m r ) = (8 + 6a 1 )m 2 + O(m 3 ). Similarly, one can obtain E { C 3 (T C) 3} C 3 [ b 6 md 6 E ( S 6 m) + 7m 3 P (T = m) 3Cb 4 md 4 E ( S 4 m) + 3C 2 b 2 mσ 2 d 2 + 3C 2 m 2 P (T = m) C 3],... (2.16) and this can be simplified along the lines of (2.16). Thus, (2.14) follows. Next, repeatedly using (2.6), we get C 3 (N C) 3 C 3 (T C) 3 + 3C 2 { (T C) 2 /C } and + 6C 2 T + C 3 (3T + 1),... (2.17) C 3 (N C) 3 C 3 (T C) 3 3C 2 (2T + 1).... (2.18) Now, the result follows by combining (2.8) - (2.9), (2.14), and (2.17) - (2.18). emma 2.5. For the stopping variable N from (1.3), we have as d 0 : E { C 2 (N C) 4} = 12σ 4 σ 4 + O(C 1 ).

9 262 nitis mukhopadhyay Proof. Following the lines of arguments as before, one obtains E { C 2 (T C) 4} C 2 [ b 8 ma 8 { m m 2 + O(m 3 ) } 4b 6 ma 6 { 1 + 6m m 2 + O(m 3 ) } +6b 4 ma 4 { 1 + 2m 1 + 2m 2 + O(m 3 ) } 4b 2 ma ] + O(m r ) = C 2 { 12m 2 + O(m 3 ) } = 12σ 4 σ 4 + O(C 1 ), and the same upper bound would similarly follow. Hence, we immediately conclude that E [ C 2 (T C) 4] = 12σ 4 σ 4 + O(C 1 ). et us denote U = C 1 2 (T C) and V = C 1 2 (N T ). From what we have shown so far, one can claim uniform integrability of U s for 0 < s 4. Next, let us write E { C 2 (N C) 4} = E(U 4 ) + 4E(U 3 V ) + 6E(U 2 V 2 ) + 4E(UV 3 ) + E(V 4 ).... (2.19) Since N T 1, we have E(V s ) = O(C s/2 ) for s = 1, 2, 3 and 4. Next, Holder s inequality leads to: E(U 3 V ) E 3 4 (U 4 )E 1 4 (V 4 ) = O(1)O(C 1 2 ) 0,... (2.20) and also E(U 3 )E(V ) = C 3 2 O(C 2 )O(C 1 2 ) 0, utilizing (2.14). Combining this with (2.20), one can claim that U 3 and V are asymptotically uncorrelated. Hence, E(U 3 V ) = O(C 1 ) asymptotically. Using similar arguments, one can see that E(U 2 V 2 ) E 2 3 ( U 3 )E 1 3 ( V 6 ) = O(1)O(C 1 ) 0, while E(U 2 )E(V 2 ) = O(1)O(C 1 ) 0, implying that U 2 and V 2 are then asymptotically uncorrelated. Thus, E(U 2 V 2 ) is asymptotically O(C 1 ). Similarly, one shows that E(UV 3 ) = O(C 1 ) as well. Hence, the result follows from (2.19), by taking expectations term by term. emma 2.6. For the stopping variable N from (1.3), we have as d 0 : E { C 5 (N C) 5} = O(C 3 ).

10 third-order approximations and beyond 263 Proof. As before, we start with E { C 5 (T C) 5} b 10 m a 10 { m m 2 + O(m 3 ) } 5b 8 ma 8 {1 + 12m 1 +56m 2 + O(m 3 ) } + 10b 6 ma 6 { 1 + 6m 1 14m 2 + O(m 3 ) } 10b 4 ma 4 { 1 + 2m 1 + 2m 2 + O(m 3 ) } + 5b 2 ma O(m r ) = { 1 + (20 + 5a 1 )m 1 + ( a a a 2 )m 2} 5 { 1 + (12 + 4a 1 )m 1 + ( a 1 + 6a a 2 )m 2} +10 { 1 + (6 + 3a 1 )m 1 + ( a 1 + 3a a 2 )m 2} 10 { 1 + (2 + 2a 1 )m 1 + (2 + 4a 1 + a a 2 )m 2} +5 { 1 + a 1 m 1 + a 2 m 2} 1 + O(m 3 ) + O(m r ) = O(C 3 ),... (2.21) and the upper bound can be tackled analogously as in previous lemmas. emma 2.7. For the stopping variable N from (1.3), we have as d 0 : E { C 3 (N C) 6} = 120σ 6 σ 6 + o(1). Proof. First, we will show that E { C 3 (T C) 6} = 120σ 6 σ 6 + O(C 1 ).... (2.22) Following the earlier lines of arguments, we get E { C 3 (T C) 6} C 3 [ b 12 m a 12 { m m m 3 + O(m 4 ) } 6b 10 m a 10 { m m m 3 + O(m 4 ) } +15b 8 ma 8 { m m m 3 + O(m 4 ) }

11 264 nitis mukhopadhyay 20b 6 ma { m m m 3 + O(m 4 ) } +15b 4 ma { m 1 + 2m 2 + 2m 3 + O(m 4 ) } 6b 2 ma ] + O(m r ) = C [{ (30 + 6a 1 )m 1 + ( a 1 + 6a a 2 1)m 2 +( a a a 2 + 6a a 1 a 2 +20a 3 1)m 3} 6 { 1 + (20 + 5a 1 )m 1 + ( a a a 2 )m 2 + ( a a a 2 + 5a a 1 a a 3 1)m 3} + 15{1+ (12 + 4a 1 )m 1 + ( a 1 + 6a a 2 )m 2 + ( a a a 2 + 4a a 1 a 2 + 4a 3 1)m 3} 20 { 1 + (6 + 3a 1 )m 1 + ( a 1 + 3a a 2 )m 2 +( a a a 2 + 3a 3 + 6a 1 a 2 + a 3 1)m 3} +15 { 1 + (2 + 2a 1 )m 1 + (2 + 4a 1 + a a 2 )m 2 +(2 + 4a 1 + 2a a 2 + 2a 3 + 2a 1 a 2 )m 3} 6 { 1 + a 1 m 1 +a 2 m 2 + a 3 m 3} O(m 4 ) ] + O(m r ) = C 3 { 120m 3 + O(m 4 ) } = 120σ 4 σ 4 + O(C 1 ),... (2.23) and the upper bound can be treated in the same fashion. Thus, (2.22) follows. Hence, uniform integrability of C 3 (T C) 6 immediately follows, since C 1 2 (T C) N(0, 2σ 2 σ 2 ). From (2.7), one notes that C 1 2 N C C 1 2 T C + 1 for C > 1, and thus C 3 (N C) 6 is uniformly integrable. In view of emma 2.2, part (i), the result then follows from (2.22). Remark 2.1. One should note that the full expansions given in (2.2) and (2.4) up to the order O(m 4 ) have been utilized only in the proof of emma 2.7. In the proofs of emmas , we had used the very same expansions

12 third-order approximations and beyond 265 terminated at the O(m 3 ) term. 3. Proofs of Theorems Recall that g(x) = 2Φ(ax 1 2 ) 1 for x > 0, defined in (1.5). Observe that, for all x > 0, we have g (1) (x) = ax 1 2 φ(ax 1 2 ), g (2) (x) = 1 2 a {x a 2 x 1 2 } φ(ax 1 2 ), g (3) (x) = 1 4 a {3x a 2 x a 4 x 1 2 } φ(ax 1 2 ), g (4) (x) = 1 8 a {15x a 2 x a 4 x a 6 x 1 2 } φ(ax 1 2 ), { } g (5) (x) = 1 16 a 105x a 2 x a 4 x a 6 x a 8 x 1 2 φ(ax 1 2 ), g (6) (x) = 1 32 a { 945x a 2 x a 4 x a 6 x a 8 x a 10 x 1 2 } φ(ax 1 2 ), with g (k) (x) = d k g(x)/dx k, k = 1,..., 6, and hence... (3.1) g (1) (1) = aφ(a), g (2) (1) = 1 2 a(a2 + 1)φ(a), g (3) (1) = 1 4 a(a4 + 2a 2 + 3)φ(a), g (4) (1) = 1 8 a(a6 + 3a 4 + 9a )φ(a), g (5) (1) = 1 16 a(a8 + 4a a a )φ(a), g (6) (1) = 1 32 a(a10 + 5a a a a )φ(a).... (3.2) Now we provide the proofs of Theorems But, first we prove the following lemma. emma 3.1. d 0 : For the stopping variable N defined in (1.3), we have as (i) (ii) E [ C 2 (N C) 4 g (4) (W ) ] = 12σ 4 σ 4 g(4) (1) + o(1); E [ C 3 (N C) 6 g (6) (W ) ] = 120σ 6 σ 6 g(6) (1) + o(1);

13 266 nitis mukhopadhyay where W is a random variable between NC 1 and 1, with derivatives g (4) ( ), g (6) ( ) defined in (3.1) - (3.2). Proof. We sketch a proof of part (i) only, since the proof of part (ii) is very similar. From (3.1), observe that g (4) (x) 4 k i x bi... (3.3) where k i s are positive numbers and b 1 = 7 2, b 2 = 5 2, b 3 = 3 2 and b 4 = 1 2. On the set {N = m}, we have W > N/C = m/c. Hence, using emma 2.1 and (3.3), [ E C 2 (N C) 4 g (4) (W )I(N = m) ] 4 i=1 k bi (m C)4 i(c/m) C P (N = m) 2 = O(1)O(m 2 )O(h 1 2 (m 1) ) = o(1).... (3.4) Next, on the set {N > m}, we have W > σ 2 σ 2. Also, C 2 (N C) 4 is uniformly integrable, and hence utilizing (3.3) again, C 2 (N C) 4 g (4) (W )I(N > m) { 4 } C 2 (N C) 4 i=1 k i(σ 2 σ 2 ) bi, so that C 2 (N C) 4 g (4) (W )I(N > m) is then uniformly integrable. Thus, in view of emma 2.2 and the fact that g (4) (x) is continuous at x = 1, we obtain [ ] E C 2 (N C) 4 g (4) (W )I(N > m) = 12σ 4 σ 4 + o(1).... (3.5) Now, (3.4) - (3.5) together complete the proof of part (i). To prove part (ii), observe from (3.1) a bound like that in (3.3) for g (6) (x) and the fact that g (6) (x) is also continuous at x = 1. Further details are left out. We keep combining terms like m s P (N = m) and replace these by a generic term O(m r ) or O(C r ) for appropriately large r(> 0). Proof of theorem 1.1. From (1.3), we note the basic inequality, i=1 b 2 ms 2 md 2 N mi(n = m) + b 2 ms 2 md which implies ( b 2 m a 2) σ 2 d 2 E(N C) ( b 2 m a 2) σ 2 d O(m r ),... (3.6) with sufficiently large r(> 0). From the expansion of b 2 ma 2 given in (2.2), it follows immediately that ( b 2 m a 2) σ 2 d 2 = a 1 Cm 1 + a 2 Cm 2 + a 3 Cm 3 + O(C 3 ) = a 1 (σ 2 /σ 2 ) + a 2(σ 2 /σ 2 )2 C 1 + a 3 (σ 2 /σ 2 )3 C 2 +O(C 3 ),

14 third-order approximations and beyond 267 and hence the result follows from (3.6). Proof of theorem 1.2. From (1.5), recall that P (µɛi N ) = E[g(N/C)], and with some random variable W between 1 and NC 1, we obtain E[g(N/C)] = g(1) + C 1 (N C)g (1) (1) C 1 (N C) 2 C g (2) (1) + 1 (N C) 3 6 C g (3) (1) C 2 (N C) 4 C g (4) (W ) (3.7) Since g (2) (1) is negative, utilizing the upper and lower bounds from emmas as well as Theorem 1.1 and emma 3.1, part (i), we get: E[g(N/C)] g(1) + { η 0 C 1 + η 1 C 2 + o(c 2 ) } g (1) (1) and also C 1 { 2 + 2σ 2 σ 2 +B 2C 1 + o(c 1 ) } g (2) (1) { 6C 1 + B 3 C 2 +o(c 2 ) } g (3) (1) C 2 { 12σ 4 σ 4 g(4) (1) + o(1) } = (1 α) + C 1 { η 0 g (1) (1) + ( σ 2 σ 2 + 1) g (2) (1) g (3) (1) } +C 2 { η 1 g (1) (1) B 2g (2) (1) B 3g (3) (1) σ4 σ 4 g4 (1) } +o(c 2 ), E[g(N/C)] g(1) + { (η 0 + 1)C 1 + η 1 C 2 + o(c 2 ) } g (1) (1) since g(1) = 1 α C 1 { 2σ 2 σ 2 2 +B 1C 1 + o(c 1 ) } g (2) (1) { 6C 1 + B 4 C 2 + o(c 2 ) } g (3) (1) C 2 { 12σ 4 σ 4 g(4) (1) + o(1) } = (1 α) + C 1 { (η 0 + 1)g (1) (1) + ( σ 2 σ 2 1) g (2) (1) + g (3) (1) } +C 2 { η 1 g (1) (1) B 1g (2) (1) B 4g (3) (1) σ4 σ 4 g4 (1) } +o(c 2 ), Proof of theorem 1.3. With some random variable W between 1 and

15 268 nitis mukhopadhyay NC 1, we can again write E[g(N/C)] = g(1) + C 1 (N C)g (1) (1) C 1 (N C) 2 C g (2) (1) + 1 (N C) 3 6 C g (3) (1) C 2 (N C) 4 C g (4) (1) 2... (3.8) + 1 (N C) C g (5) (1) C 3 (N C) 6 C g (6) (W ). 3 Recall that g (2) (1), g (4) (1) are both negative. Again utilizing the upper and lower bounds from emmas as well as Theorem 1.1 and emma 3.1, part (ii), we get: E[g(N/C)] g(1) + { η 0 C 1 + η 1 C 2 + O(C 3 ) } g (1) (1) C 1 { 2 + 2σ 2 σ 2 + B 2C 1 +O(C 2 ) } g (2) (1) { 6C 1 + B 3 C 2 + O(C 3 ) } g (3) (1) C 2 { 12σ 4 σ 4 + O(C 1 ) } g (4) (1) + O(C 3 ) +O(C 3 ) { 120σ 6 σ 6 + o(1)} = (1 α) + C 1 { η 0 g (1) (1) + ( σ 2 σ 2 + 1) g (2) (1) g (3) (1) } +C 2 { η 1 g (1) (1) B 2g (2) (1) B 4g (3) (1) σ4 σ 4 g(4) (1) } +O(C 3 ), and the upper bound is similar. Remark 3.1. In order to replace o(d 4 ) order for the remainder term in Theorem 1.2 by the significantly sharper rate O(d 6 ), the emmas made all the difference. Remark 3.2. Mukhopadhyay and Duggan (1997) had included some simulation results with µ = 5, σ = 5, σ = 2, 4, m 0 = 10 and C = 30, 40, 50, 100, 200, 300, 400, 500, 800, based on 5000 independent replications for each entry. Their second-order interval for E(N C) was taken to be (η 0, η 0 +1), and it was found that (n C) values were largely within this interval while, for the most part, staying closer to η 0 than (η 0 +1) for larger values of C(> 200). For C = 30, 40, 50 and 100, the corresponding (n C) values stayed within (η, η + 1) where η = η 0 +η 1 C 1 +η 2 C 2, given by Theorem 1.1, and usually staying much closer to η. The usefulness of the third- and higher order approximations, particularly for small and moderate values of C, is demonstrably clear.

16 third-order approximations and beyond 269 Acknowledgement. The author thanks Professor Bimal K. Sinha for his help in preparing a clearer presentation. References Chow, Y.S. and Robbins, H. (1965). On the asymptotic theory of fixed-width sequential confidence intervals for the mean. Ann. Math. Statist. 36, Cohen, A. and Sackrowitz, H.B. (1984). Bayes double sampling estimation procedures. Ann. Statist. 12, Dantzig, G.B. (1940). On the nonexistence of tests of Student s hypothesis having power functions independent of σ. Ann. Math. Statist. 11, Ghosh, M. and Mukhopadhyay, N. (1981). Consistency and asymptotic efficiency of twostage and sequential estimation procedures. Sankhya Ser. A 43, Ghosh, M., Mukhopadhyay, N. and Sen, P.K. (1997). Sequential Estimation. John Wiley & Sons., Inc., New York. Johnson, N.. and Kotz, S. (1970). Continuous Univariate Distributions 2. John Wiley & Sons., Inc., New York. Mukhopadhyay, N. (1980). A consistent and asymptotically efficient two-stage procedure to construct fixed-width confidence intervals for the mean. Metrika 27, Mukhopadhyay, N. and Duggan, W.T. (1997). Can a two-stage procedure enjoy secondorder properties? Sankhya Ser. A 59, Ray, W.D. (1957). Sequential confidence intervals for the mean of a normal population with unknown variance. J. Roy. Statist. Soc. Ser. B 19, Stein, C. (1945). A two sample test for a linear hypothesis whose power is independent of the variance. Ann. Math. Statist. 16, (1949). Some problems in sequential estimation (abstract). Econometrica 17, Woodroofe, M. (1977). Second-order approximations for sequential point and interval estimation. Ann. Statist. 5, N. Mukhopadhyay Department of Statistics University of Connecticut Storrs CT mukhop@uconnvm.uconn.edu

CONVERGENCE RATES FOR rwo-stage CONFIDENCE INTERVALS BASED ON U-STATISTICS

CONVERGENCE RATES FOR rwo-stage CONFIDENCE INTERVALS BASED ON U-STATISTICS Ann. Inst. Statist. Math. Vol. 40, No. 1, 111-117 (1988) CONVERGENCE RATES FOR rwo-stage CONFIDENCE INTERVALS BASED ON U-STATISTICS N. MUKHOPADHYAY 1 AND G. VIK 2 i Department of Statistics, University

More information

Confidence Intervals of Prescribed Precision Summary

Confidence Intervals of Prescribed Precision Summary Confidence Intervals of Prescribed Precision Summary Charles Stein showed in 1945 that by using a two stage sequential procedure one could give a confidence interval for the mean of a normal distribution

More information

FIXED-WIDTH CONFIDENCE INTERVAL FOR A LOGNORMAL MEAN

FIXED-WIDTH CONFIDENCE INTERVAL FOR A LOGNORMAL MEAN IJMMS 9:3 00 143 153 PII. S01611710011055 http://ijmms.hindawi.com Hindawi Publishing Corp. FIXED-WIDTH CONFIDENCE INTERVAL FOR A LOGNORMAL MEAN MAKOTO AOSHIMA and ZAKKULA GOVINDARAJULU Received 5 October

More information

Sequential Procedure for Testing Hypothesis about Mean of Latent Gaussian Process

Sequential Procedure for Testing Hypothesis about Mean of Latent Gaussian Process Applied Mathematical Sciences, Vol. 4, 2010, no. 62, 3083-3093 Sequential Procedure for Testing Hypothesis about Mean of Latent Gaussian Process Julia Bondarenko Helmut-Schmidt University Hamburg University

More information

High-dimensional asymptotic expansions for the distributions of canonical correlations

High-dimensional asymptotic expansions for the distributions of canonical correlations Journal of Multivariate Analysis 100 2009) 231 242 Contents lists available at ScienceDirect Journal of Multivariate Analysis journal homepage: www.elsevier.com/locate/jmva High-dimensional asymptotic

More information

A NOTE ON A DISTRIBUTION OF WEIGHTED SUMS OF I.I.D. RAYLEIGH RANDOM VARIABLES

A NOTE ON A DISTRIBUTION OF WEIGHTED SUMS OF I.I.D. RAYLEIGH RANDOM VARIABLES Sankhyā : The Indian Journal of Statistics 1998, Volume 6, Series A, Pt. 2, pp. 171-175 A NOTE ON A DISTRIBUTION OF WEIGHTED SUMS OF I.I.D. RAYLEIGH RANDOM VARIABLES By P. HITCZENKO North Carolina State

More information

Multistage Methodologies for Partitioning a Set of Exponential. populations.

Multistage Methodologies for Partitioning a Set of Exponential. populations. Multistage Methodologies for Partitioning a Set of Exponential Populations Department of Mathematics, University of New Orleans, 2000 Lakefront, New Orleans, LA 70148, USA tsolanky@uno.edu Tumulesh K.

More information

Multistage Sampling Strategies and Inference in Health Studies Under Appropriate Linex Loss Functions

Multistage Sampling Strategies and Inference in Health Studies Under Appropriate Linex Loss Functions University of Connecticut DigitalCommons@UConn Doctoral Dissertations University of Connecticut Graduate School 7-7-207 Multistage Sampling Strategies and Inference in Health Studies Under Appropriate

More information

NUMBER FIELDS WITHOUT SMALL GENERATORS

NUMBER FIELDS WITHOUT SMALL GENERATORS NUMBER FIELDS WITHOUT SMALL GENERATORS JEFFREY D. VAALER AND MARTIN WIDMER Abstract. Let D > be an integer, and let b = b(d) > be its smallest divisor. We show that there are infinitely many number fields

More information

MOMENT CONVERGENCE RATES OF LIL FOR NEGATIVELY ASSOCIATED SEQUENCES

MOMENT CONVERGENCE RATES OF LIL FOR NEGATIVELY ASSOCIATED SEQUENCES J. Korean Math. Soc. 47 1, No., pp. 63 75 DOI 1.4134/JKMS.1.47..63 MOMENT CONVERGENCE RATES OF LIL FOR NEGATIVELY ASSOCIATED SEQUENCES Ke-Ang Fu Li-Hua Hu Abstract. Let X n ; n 1 be a strictly stationary

More information

Estimation of parametric functions in Downton s bivariate exponential distribution

Estimation of parametric functions in Downton s bivariate exponential distribution Estimation of parametric functions in Downton s bivariate exponential distribution George Iliopoulos Department of Mathematics University of the Aegean 83200 Karlovasi, Samos, Greece e-mail: geh@aegean.gr

More information

A sequential hypothesis test based on a generalized Azuma inequality 1

A sequential hypothesis test based on a generalized Azuma inequality 1 A sequential hypothesis test based on a generalized Azuma inequality 1 Daniël Reijsbergen a,2, Werner Scheinhardt b, Pieter-Tjerk de Boer b a Laboratory for Foundations of Computer Science, University

More information

Asymptotic efficiency of simple decisions for the compound decision problem

Asymptotic efficiency of simple decisions for the compound decision problem Asymptotic efficiency of simple decisions for the compound decision problem Eitan Greenshtein and Ya acov Ritov Department of Statistical Sciences Duke University Durham, NC 27708-0251, USA e-mail: eitan.greenshtein@gmail.com

More information

More Powerful Tests for Homogeneity of Multivariate Normal Mean Vectors under an Order Restriction

More Powerful Tests for Homogeneity of Multivariate Normal Mean Vectors under an Order Restriction Sankhyā : The Indian Journal of Statistics 2007, Volume 69, Part 4, pp. 700-716 c 2007, Indian Statistical Institute More Powerful Tests for Homogeneity of Multivariate Normal Mean Vectors under an Order

More information

A general sequential fixed-accuracy confidence interval estimation methodology for a positive parameter: illustrations using health and safety data

A general sequential fixed-accuracy confidence interval estimation methodology for a positive parameter: illustrations using health and safety data Ann Inst Stat Math (2016) 68:541 570 DOI 10.1007/s10463-015-0504-2 A general sequential fixed-accuracy confidence interval estimation methodology for a positive parameter: illustrations using health and

More information

Testing Simple Hypotheses R.L. Wolpert Institute of Statistics and Decision Sciences Duke University, Box Durham, NC 27708, USA

Testing Simple Hypotheses R.L. Wolpert Institute of Statistics and Decision Sciences Duke University, Box Durham, NC 27708, USA Testing Simple Hypotheses R.L. Wolpert Institute of Statistics and Decision Sciences Duke University, Box 90251 Durham, NC 27708, USA Summary: Pre-experimental Frequentist error probabilities do not summarize

More information

FIXED-WIDTH CONFIDENCE INTERVALS FOR CONTRASTS IN THE MEANS AJIT CHATURVEDI 1, N. D. SHUKLA 2 AND PRAMOD S. SHUKLA 2

FIXED-WIDTH CONFIDENCE INTERVALS FOR CONTRASTS IN THE MEANS AJIT CHATURVEDI 1, N. D. SHUKLA 2 AND PRAMOD S. SHUKLA 2 Ann. Inst. Statist. Math. Vol. 44, No. 1, 157-167 (1992) FIXED-WIDTH CONFIDENCE INTERVALS FOR CONTRASTS IN THE MEANS AJIT CHATURVEDI 1, N. D. SHUKLA 2 AND PRAMOD S. SHUKLA 2 1 Department of Statistics,

More information

A NONINFORMATIVE BAYESIAN APPROACH FOR TWO-STAGE CLUSTER SAMPLING

A NONINFORMATIVE BAYESIAN APPROACH FOR TWO-STAGE CLUSTER SAMPLING Sankhyā : The Indian Journal of Statistics Special Issue on Sample Surveys 1999, Volume 61, Series B, Pt. 1, pp. 133-144 A OIFORMATIVE BAYESIA APPROACH FOR TWO-STAGE CLUSTER SAMPLIG By GLE MEEDE University

More information

FIXED-WIDTH CONFIDENCE INTERVAL FOR A LOGNORMAL MEAN

FIXED-WIDTH CONFIDENCE INTERVAL FOR A LOGNORMAL MEAN IJMMS 9:3 00 143 153 PII. S01611710011055 http://ijmms.hindawi.com Hindawi Publishing Corp. FIXED-WIDTH CONFIDENCE INTERVAL FOR A LOGNORMAL MEAN MAKOTO AOSHIMA and ZAKKULA GOVINDARAJULU Received 5 October

More information

Applying the Benjamini Hochberg procedure to a set of generalized p-values

Applying the Benjamini Hochberg procedure to a set of generalized p-values U.U.D.M. Report 20:22 Applying the Benjamini Hochberg procedure to a set of generalized p-values Fredrik Jonsson Department of Mathematics Uppsala University Applying the Benjamini Hochberg procedure

More information

Submitted to the Brazilian Journal of Probability and Statistics

Submitted to the Brazilian Journal of Probability and Statistics Submitted to the Brazilian Journal of Probability and Statistics Multivariate normal approximation of the maximum likelihood estimator via the delta method Andreas Anastasiou a and Robert E. Gaunt b a

More information

Studentization and Prediction in a Multivariate Normal Setting

Studentization and Prediction in a Multivariate Normal Setting Studentization and Prediction in a Multivariate Normal Setting Morris L. Eaton University of Minnesota School of Statistics 33 Ford Hall 4 Church Street S.E. Minneapolis, MN 55455 USA eaton@stat.umn.edu

More information

A Note on Hypothesis Testing with Random Sample Sizes and its Relationship to Bayes Factors

A Note on Hypothesis Testing with Random Sample Sizes and its Relationship to Bayes Factors Journal of Data Science 6(008), 75-87 A Note on Hypothesis Testing with Random Sample Sizes and its Relationship to Bayes Factors Scott Berry 1 and Kert Viele 1 Berry Consultants and University of Kentucky

More information

Mean. Pranab K. Mitra and Bimal K. Sinha. Department of Mathematics and Statistics, University Of Maryland, Baltimore County

Mean. Pranab K. Mitra and Bimal K. Sinha. Department of Mathematics and Statistics, University Of Maryland, Baltimore County A Generalized p-value Approach to Inference on Common Mean Pranab K. Mitra and Bimal K. Sinha Department of Mathematics and Statistics, University Of Maryland, Baltimore County 1000 Hilltop Circle, Baltimore,

More information

Non Uniform Bounds on Geometric Approximation Via Stein s Method and w-functions

Non Uniform Bounds on Geometric Approximation Via Stein s Method and w-functions Communications in Statistics Theory and Methods, 40: 45 58, 20 Copyright Taylor & Francis Group, LLC ISSN: 036-0926 print/532-45x online DOI: 0.080/036092090337778 Non Uniform Bounds on Geometric Approximation

More information

STEIN-TYPE IMPROVEMENTS OF CONFIDENCE INTERVALS FOR THE GENERALIZED VARIANCE

STEIN-TYPE IMPROVEMENTS OF CONFIDENCE INTERVALS FOR THE GENERALIZED VARIANCE Ann. Inst. Statist. Math. Vol. 43, No. 2, 369-375 (1991) STEIN-TYPE IMPROVEMENTS OF CONFIDENCE INTERVALS FOR THE GENERALIZED VARIANCE SANAT K. SARKAR Department of Statistics, Temple University, Philadelphia,

More information

Primer on statistics:

Primer on statistics: Primer on statistics: MLE, Confidence Intervals, and Hypothesis Testing ryan.reece@gmail.com http://rreece.github.io/ Insight Data Science - AI Fellows Workshop Feb 16, 018 Outline 1. Maximum likelihood

More information

Wald for non-stopping times: The rewards of impatient prophets

Wald for non-stopping times: The rewards of impatient prophets Electron. Commun. Probab. 19 (2014), no. 78, 1 9. DOI: 10.1214/ECP.v19-3609 ISSN: 1083-589X ELECTRONIC COMMUNICATIONS in PROBABILITY Wald for non-stopping times: The rewards of impatient prophets Alexander

More information

APPROXIMATION OF ENTIRE FUNCTIONS OF SLOW GROWTH ON COMPACT SETS. G. S. Srivastava and Susheel Kumar

APPROXIMATION OF ENTIRE FUNCTIONS OF SLOW GROWTH ON COMPACT SETS. G. S. Srivastava and Susheel Kumar ARCHIVUM MATHEMATICUM BRNO) Tomus 45 2009), 137 146 APPROXIMATION OF ENTIRE FUNCTIONS OF SLOW GROWTH ON COMPACT SETS G. S. Srivastava and Susheel Kumar Abstract. In the present paper, we study the polynomial

More information

Packing and decomposition of graphs with trees

Packing and decomposition of graphs with trees Packing and decomposition of graphs with trees Raphael Yuster Department of Mathematics University of Haifa-ORANIM Tivon 36006, Israel. e-mail: raphy@math.tau.ac.il Abstract Let H be a tree on h 2 vertices.

More information

= 1 2x. x 2 a ) 0 (mod p n ), (x 2 + 2a + a2. x a ) 2

= 1 2x. x 2 a ) 0 (mod p n ), (x 2 + 2a + a2. x a ) 2 8. p-adic numbers 8.1. Motivation: Solving x 2 a (mod p n ). Take an odd prime p, and ( an) integer a coprime to p. Then, as we know, x 2 a (mod p) has a solution x Z iff = 1. In this case we can suppose

More information

Appendix B for The Evolution of Strategic Sophistication (Intended for Online Publication)

Appendix B for The Evolution of Strategic Sophistication (Intended for Online Publication) Appendix B for The Evolution of Strategic Sophistication (Intended for Online Publication) Nikolaus Robalino and Arthur Robson Appendix B: Proof of Theorem 2 This appendix contains the proof of Theorem

More information

The concentration of the chromatic number of random graphs

The concentration of the chromatic number of random graphs The concentration of the chromatic number of random graphs Noga Alon Michael Krivelevich Abstract We prove that for every constant δ > 0 the chromatic number of the random graph G(n, p) with p = n 1/2

More information

Consistency of Test-based Criterion for Selection of Variables in High-dimensional Two Group-Discriminant Analysis

Consistency of Test-based Criterion for Selection of Variables in High-dimensional Two Group-Discriminant Analysis Consistency of Test-based Criterion for Selection of Variables in High-dimensional Two Group-Discriminant Analysis Yasunori Fujikoshi and Tetsuro Sakurai Department of Mathematics, Graduate School of Science,

More information

Lines With Many Points On Both Sides

Lines With Many Points On Both Sides Lines With Many Points On Both Sides Rom Pinchasi Hebrew University of Jerusalem and Massachusetts Institute of Technology September 13, 2002 Abstract Let G be a finite set of points in the plane. A line

More information

Yunhi Cho and Young-One Kim

Yunhi Cho and Young-One Kim Bull. Korean Math. Soc. 41 (2004), No. 1, pp. 27 43 ANALYTIC PROPERTIES OF THE LIMITS OF THE EVEN AND ODD HYPERPOWER SEQUENCES Yunhi Cho Young-One Kim Dedicated to the memory of the late professor Eulyong

More information

2tdt 1 y = t2 + C y = which implies C = 1 and the solution is y = 1

2tdt 1 y = t2 + C y = which implies C = 1 and the solution is y = 1 Lectures - Week 11 General First Order ODEs & Numerical Methods for IVPs In general, nonlinear problems are much more difficult to solve than linear ones. Unfortunately many phenomena exhibit nonlinear

More information

ON THE COMPLETE CONVERGENCE FOR WEIGHTED SUMS OF DEPENDENT RANDOM VARIABLES UNDER CONDITION OF WEIGHTED INTEGRABILITY

ON THE COMPLETE CONVERGENCE FOR WEIGHTED SUMS OF DEPENDENT RANDOM VARIABLES UNDER CONDITION OF WEIGHTED INTEGRABILITY J. Korean Math. Soc. 45 (2008), No. 4, pp. 1101 1111 ON THE COMPLETE CONVERGENCE FOR WEIGHTED SUMS OF DEPENDENT RANDOM VARIABLES UNDER CONDITION OF WEIGHTED INTEGRABILITY Jong-Il Baek, Mi-Hwa Ko, and Tae-Sung

More information

Notes 18 : Optional Sampling Theorem

Notes 18 : Optional Sampling Theorem Notes 18 : Optional Sampling Theorem Math 733-734: Theory of Probability Lecturer: Sebastien Roch References: [Wil91, Chapter 14], [Dur10, Section 5.7]. Recall: DEF 18.1 (Uniform Integrability) A collection

More information

Consistency of test based method for selection of variables in high dimensional two group discriminant analysis

Consistency of test based method for selection of variables in high dimensional two group discriminant analysis https://doi.org/10.1007/s42081-019-00032-4 ORIGINAL PAPER Consistency of test based method for selection of variables in high dimensional two group discriminant analysis Yasunori Fujikoshi 1 Tetsuro Sakurai

More information

THE CLASSIFICATION OF PLANAR MONOMIALS OVER FIELDS OF PRIME SQUARE ORDER

THE CLASSIFICATION OF PLANAR MONOMIALS OVER FIELDS OF PRIME SQUARE ORDER THE CLASSIFICATION OF PLANAR MONOMIALS OVER FIELDS OF PRIME SQUARE ORDER ROBERT S COULTER Abstract Planar functions were introduced by Dembowski and Ostrom in [3] to describe affine planes possessing collineation

More information

Some History of Optimality

Some History of Optimality IMS Lecture Notes- Monograph Series Optimality: The Third Erich L. Lehmann Symposium Vol. 57 (2009) 11-17 @ Institute of Mathematical Statistics, 2009 DOl: 10.1214/09-LNMS5703 Erich L. Lehmann University

More information

Bahadur representations for bootstrap quantiles 1

Bahadur representations for bootstrap quantiles 1 Bahadur representations for bootstrap quantiles 1 Yijun Zuo Department of Statistics and Probability, Michigan State University East Lansing, MI 48824, USA zuo@msu.edu 1 Research partially supported by

More information

SELECTING THE NORMAL POPULATION WITH THE SMALLEST COEFFICIENT OF VARIATION

SELECTING THE NORMAL POPULATION WITH THE SMALLEST COEFFICIENT OF VARIATION SELECTING THE NORMAL POPULATION WITH THE SMALLEST COEFFICIENT OF VARIATION Ajit C. Tamhane Department of IE/MS and Department of Statistics Northwestern University, Evanston, IL 60208 Anthony J. Hayter

More information

arxiv: v1 [math.st] 26 Jun 2011

arxiv: v1 [math.st] 26 Jun 2011 The Shape of the Noncentral χ 2 Density arxiv:1106.5241v1 [math.st] 26 Jun 2011 Yaming Yu Department of Statistics University of California Irvine, CA 92697, USA yamingy@uci.edu Abstract A noncentral χ

More information

On the exact computation of the density and of the quantiles of linear combinations of t and F random variables

On the exact computation of the density and of the quantiles of linear combinations of t and F random variables Journal of Statistical Planning and Inference 94 3 www.elsevier.com/locate/jspi On the exact computation of the density and of the quantiles of linear combinations of t and F random variables Viktor Witkovsky

More information

P(I -ni < an for all n > in) = 1 - Pm# 1

P(I -ni < an for all n > in) = 1 - Pm# 1 ITERATED LOGARITHM INEQUALITIES* By D. A. DARLING AND HERBERT ROBBINS UNIVERSITY OF CALIFORNIA, BERKELEY Communicated by J. Neyman, March 10, 1967 1. Introduction.-Let x,x1,x2,... be a sequence of independent,

More information

k-protected VERTICES IN BINARY SEARCH TREES

k-protected VERTICES IN BINARY SEARCH TREES k-protected VERTICES IN BINARY SEARCH TREES MIKLÓS BÓNA Abstract. We show that for every k, the probability that a randomly selected vertex of a random binary search tree on n nodes is at distance k from

More information

University of California, Berkeley

University of California, Berkeley University of California, Berkeley U.C. Berkeley Division of Biostatistics Working Paper Series Year 24 Paper 153 A Note on Empirical Likelihood Inference of Residual Life Regression Ying Qing Chen Yichuan

More information

An Asymptotic Formula for Goldbach s Conjecture with Monic Polynomials in Z[x]

An Asymptotic Formula for Goldbach s Conjecture with Monic Polynomials in Z[x] An Asymptotic Formula for Goldbach s Conjecture with Monic Polynomials in Z[x] Mark Kozek 1 Introduction. In a recent Monthly note, Saidak [6], improving on a result of Hayes [1], gave Chebyshev-type estimates

More information

Example continued. Math 425 Intro to Probability Lecture 37. Example continued. Example

Example continued. Math 425 Intro to Probability Lecture 37. Example continued. Example continued : Coin tossing Math 425 Intro to Probability Lecture 37 Kenneth Harris kaharri@umich.edu Department of Mathematics University of Michigan April 8, 2009 Consider a Bernoulli trials process with

More information

A LIMIT THEOREM FOR THE IMBALANCE OF A ROTATING WHEEL

A LIMIT THEOREM FOR THE IMBALANCE OF A ROTATING WHEEL Sankhyā : The Indian Journal of Statistics 1994, Volume 56, Series B, Pt. 3, pp. 46-467 A LIMIT THEOREM FOR THE IMBALANCE OF A ROTATING WHEEL By K.G. RAMAMURTHY and V. RAJENDRA PRASAD Indian Statistical

More information

A NOTE ON THE COMPLETE MOMENT CONVERGENCE FOR ARRAYS OF B-VALUED RANDOM VARIABLES

A NOTE ON THE COMPLETE MOMENT CONVERGENCE FOR ARRAYS OF B-VALUED RANDOM VARIABLES Bull. Korean Math. Soc. 52 (205), No. 3, pp. 825 836 http://dx.doi.org/0.434/bkms.205.52.3.825 A NOTE ON THE COMPLETE MOMENT CONVERGENCE FOR ARRAYS OF B-VALUED RANDOM VARIABLES Yongfeng Wu and Mingzhu

More information

Asymptotic Nonequivalence of Nonparametric Experiments When the Smoothness Index is ½

Asymptotic Nonequivalence of Nonparametric Experiments When the Smoothness Index is ½ University of Pennsylvania ScholarlyCommons Statistics Papers Wharton Faculty Research 1998 Asymptotic Nonequivalence of Nonparametric Experiments When the Smoothness Index is ½ Lawrence D. Brown University

More information

Daniel M. Oberlin Department of Mathematics, Florida State University. January 2005

Daniel M. Oberlin Department of Mathematics, Florida State University. January 2005 PACKING SPHERES AND FRACTAL STRICHARTZ ESTIMATES IN R d FOR d 3 Daniel M. Oberlin Department of Mathematics, Florida State University January 005 Fix a dimension d and for x R d and r > 0, let Sx, r) stand

More information

THE ANALYSIS OF SOLUTION OF NONLINEAR SECOND ORDER ORDINARY DIFFERENTIAL EQUATIONS

THE ANALYSIS OF SOLUTION OF NONLINEAR SECOND ORDER ORDINARY DIFFERENTIAL EQUATIONS THE ANALYSIS OF SOLUTION OF NONLINEAR SECOND ORDER ORDINARY DIFFERENTIAL EQUATIONS FANIRAN TAYE SAMUEL Assistant Lecturer, Department of Computer Science, Lead City University, Ibadan, Nigeria. Email :

More information

Optimal Sequential Procedures with Bayes Decision Rules

Optimal Sequential Procedures with Bayes Decision Rules International Mathematical Forum, 5, 2010, no. 43, 2137-2147 Optimal Sequential Procedures with Bayes Decision Rules Andrey Novikov Department of Mathematics Autonomous Metropolitan University - Iztapalapa

More information

A GENERAL FRAMEWORK FOR THE ASYMPTOTIC VALIDITY OF TWO-STAGE PROCEDURES FOR SELECTION AND MULTIPLE COMPARISONS WITH CONSISTENT VARIANCE ESTIMATORS

A GENERAL FRAMEWORK FOR THE ASYMPTOTIC VALIDITY OF TWO-STAGE PROCEDURES FOR SELECTION AND MULTIPLE COMPARISONS WITH CONSISTENT VARIANCE ESTIMATORS Proceedings of the 2009 Winter Simulation Conference M. D. Rossetti, R. R. Hill, B. Johansson, A. Dunkin, and R. G. Ingalls, eds. A GENERAL FRAMEWORK FOR THE ASYMPTOTIC VALIDITY OF TWO-STAGE PROCEDURES

More information

On the Average Path Length of Complete m-ary Trees

On the Average Path Length of Complete m-ary Trees 1 2 3 47 6 23 11 Journal of Integer Sequences, Vol. 17 2014, Article 14.6.3 On the Average Path Length of Complete m-ary Trees M. A. Nyblom School of Mathematics and Geospatial Science RMIT University

More information

STATISTICS SYLLABUS UNIT I

STATISTICS SYLLABUS UNIT I STATISTICS SYLLABUS UNIT I (Probability Theory) Definition Classical and axiomatic approaches.laws of total and compound probability, conditional probability, Bayes Theorem. Random variable and its distribution

More information

A NOTE ON QUASI ISOMETRIES II. S.M. Patel Sardar Patel University, India

A NOTE ON QUASI ISOMETRIES II. S.M. Patel Sardar Patel University, India GLASNIK MATEMATIČKI Vol. 38(58)(2003), 111 120 A NOTE ON QUASI ISOMETRIES II S.M. Patel Sardar Patel University, India Abstract. An operator A on a complex Hilbert space H is called a quasi-isometry if

More information

Bayesian decision theory Introduction to Pattern Recognition. Lectures 4 and 5: Bayesian decision theory

Bayesian decision theory Introduction to Pattern Recognition. Lectures 4 and 5: Bayesian decision theory Bayesian decision theory 8001652 Introduction to Pattern Recognition. Lectures 4 and 5: Bayesian decision theory Jussi Tohka jussi.tohka@tut.fi Institute of Signal Processing Tampere University of Technology

More information

LARGE DEVIATIONS OF TYPICAL LINEAR FUNCTIONALS ON A CONVEX BODY WITH UNCONDITIONAL BASIS. S. G. Bobkov and F. L. Nazarov. September 25, 2011

LARGE DEVIATIONS OF TYPICAL LINEAR FUNCTIONALS ON A CONVEX BODY WITH UNCONDITIONAL BASIS. S. G. Bobkov and F. L. Nazarov. September 25, 2011 LARGE DEVIATIONS OF TYPICAL LINEAR FUNCTIONALS ON A CONVEX BODY WITH UNCONDITIONAL BASIS S. G. Bobkov and F. L. Nazarov September 25, 20 Abstract We study large deviations of linear functionals on an isotropic

More information

CYCLES AND FIXED POINTS OF HAPPY FUNCTIONS

CYCLES AND FIXED POINTS OF HAPPY FUNCTIONS Journal of Combinatorics and Number Theory Volume 2, Issue 3, pp. 65-77 ISSN 1942-5600 c 2010 Nova Science Publishers, Inc. CYCLES AND FIXED POINTS OF HAPPY FUNCTIONS Kathryn Hargreaves and Samir Siksek

More information

δ -method and M-estimation

δ -method and M-estimation Econ 2110, fall 2016, Part IVb Asymptotic Theory: δ -method and M-estimation Maximilian Kasy Department of Economics, Harvard University 1 / 40 Example Suppose we estimate the average effect of class size

More information

Multistage Tests of Multiple Hypotheses

Multistage Tests of Multiple Hypotheses Communications in Statistics Theory and Methods, 39: 1597 167, 21 Copyright Taylor & Francis Group, LLC ISSN: 361-926 print/1532-415x online DOI: 1.18/3619282592852 Multistage Tests of Multiple Hypotheses

More information

BURGESS INEQUALITY IN F p 2. Mei-Chu Chang

BURGESS INEQUALITY IN F p 2. Mei-Chu Chang BURGESS INEQUALITY IN F p 2 Mei-Chu Chang Abstract. Let be a nontrivial multiplicative character of F p 2. We obtain the following results.. Given ε > 0, there is δ > 0 such that if ω F p 2\F p and I,

More information

Recall that if X 1,...,X n are random variables with finite expectations, then. The X i can be continuous or discrete or of any other type.

Recall that if X 1,...,X n are random variables with finite expectations, then. The X i can be continuous or discrete or of any other type. Expectations of Sums of Random Variables STAT/MTHE 353: 4 - More on Expectations and Variances T. Linder Queen s University Winter 017 Recall that if X 1,...,X n are random variables with finite expectations,

More information

MAS223 Statistical Inference and Modelling Exercises

MAS223 Statistical Inference and Modelling Exercises MAS223 Statistical Inference and Modelling Exercises The exercises are grouped into sections, corresponding to chapters of the lecture notes Within each section exercises are divided into warm-up questions,

More information

Two-Stage and Sequential Estimation of Parameter N of Binomial Distribution When p Is Known

Two-Stage and Sequential Estimation of Parameter N of Binomial Distribution When p Is Known Two-Stage and Sequential Estimation of Parameter N of Binomial Distribution When p Is Known Shyamal K. De 1 and Shelemyahu Zacks 2 1 School of Mathematical Sciences, National Institute of Science Education

More information

A NUMBER-THEORETIC CONJECTURE AND ITS IMPLICATION FOR SET THEORY. 1. Motivation

A NUMBER-THEORETIC CONJECTURE AND ITS IMPLICATION FOR SET THEORY. 1. Motivation Acta Math. Univ. Comenianae Vol. LXXIV, 2(2005), pp. 243 254 243 A NUMBER-THEORETIC CONJECTURE AND ITS IMPLICATION FOR SET THEORY L. HALBEISEN Abstract. For any set S let seq (S) denote the cardinality

More information

A Readable Introduction to Real Mathematics

A Readable Introduction to Real Mathematics Solutions to selected problems in the book A Readable Introduction to Real Mathematics D. Rosenthal, D. Rosenthal, P. Rosenthal Chapter 7: The Euclidean Algorithm and Applications 1. Find the greatest

More information

PRIME NUMBERS YANKI LEKILI

PRIME NUMBERS YANKI LEKILI PRIME NUMBERS YANKI LEKILI We denote by N the set of natural numbers: 1,2,..., These are constructed using Peano axioms. We will not get into the philosophical questions related to this and simply assume

More information

T 2 Type Test Statistic and Simultaneous Confidence Intervals for Sub-mean Vectors in k-sample Problem

T 2 Type Test Statistic and Simultaneous Confidence Intervals for Sub-mean Vectors in k-sample Problem T Type Test Statistic and Simultaneous Confidence Intervals for Sub-mean Vectors in k-sample Problem Toshiki aito a, Tamae Kawasaki b and Takashi Seo b a Department of Applied Mathematics, Graduate School

More information

2 ANDREW L. RUKHIN We accept here the usual in the change-point analysis convention that, when the minimizer in ) is not determined uniquely, the smal

2 ANDREW L. RUKHIN We accept here the usual in the change-point analysis convention that, when the minimizer in ) is not determined uniquely, the smal THE RATES OF CONVERGENCE OF BAYES ESTIMATORS IN CHANGE-OINT ANALYSIS ANDREW L. RUKHIN Department of Mathematics and Statistics UMBC Baltimore, MD 2228 USA Abstract In the asymptotic setting of the change-point

More information

An almost sure invariance principle for additive functionals of Markov chains

An almost sure invariance principle for additive functionals of Markov chains Statistics and Probability Letters 78 2008 854 860 www.elsevier.com/locate/stapro An almost sure invariance principle for additive functionals of Markov chains F. Rassoul-Agha a, T. Seppäläinen b, a Department

More information

Expansion properties of a random regular graph after random vertex deletions

Expansion properties of a random regular graph after random vertex deletions Expansion properties of a random regular graph after random vertex deletions Catherine Greenhill School of Mathematics and Statistics The University of New South Wales Sydney NSW 05, Australia csg@unsw.edu.au

More information

Testing Statistical Hypotheses

Testing Statistical Hypotheses E.L. Lehmann Joseph P. Romano Testing Statistical Hypotheses Third Edition 4y Springer Preface vii I Small-Sample Theory 1 1 The General Decision Problem 3 1.1 Statistical Inference and Statistical Decisions

More information

New Lower Bounds for the Number of Blocks in Balanced Incomplete Block Designs

New Lower Bounds for the Number of Blocks in Balanced Incomplete Block Designs New Lower Bounds for the Number of Blocs in Balanced Incomplete Bloc Designs MUHAMMAD A KHAN ABSTRACT: Bose [1] proved the inequality b v+ r 1 for resolvable balanced incomplete bloc designs (RBIBDs) and

More information

NOTES ON ZHANG S PRIME GAPS PAPER

NOTES ON ZHANG S PRIME GAPS PAPER NOTES ON ZHANG S PRIME GAPS PAPER TERENCE TAO. Zhang s results For any natural number H, let P (H) denote the assertion that there are infinitely many pairs of distinct primes p, q with p q H; thus for

More information

Confidence Regions For The Ratio Of Two Percentiles

Confidence Regions For The Ratio Of Two Percentiles Confidence Regions For The Ratio Of Two Percentiles Richard Johnson Joint work with Li-Fei Huang and Songyong Sim January 28, 2009 OUTLINE Introduction Exact sampling results normal linear model case Other

More information

Forcing unbalanced complete bipartite minors

Forcing unbalanced complete bipartite minors Forcing unbalanced complete bipartite minors Daniela Kühn Deryk Osthus Abstract Myers conjectured that for every integer s there exists a positive constant C such that for all integers t every graph of

More information

Non-separating 2-factors of an even-regular graph

Non-separating 2-factors of an even-regular graph Discrete Mathematics 308 008) 5538 5547 www.elsevier.com/locate/disc Non-separating -factors of an even-regular graph Yusuke Higuchi a Yui Nomura b a Mathematics Laboratories College of Arts and Sciences

More information

Conway s RATS Sequences in Base 3

Conway s RATS Sequences in Base 3 3 47 6 3 Journal of Integer Sequences, Vol. 5 (0), Article.9. Conway s RATS Sequences in Base 3 Johann Thiel Department of Mathematical Sciences United States Military Academy West Point, NY 0996 USA johann.thiel@usma.edu

More information

arxiv: v5 [math.na] 16 Nov 2017

arxiv: v5 [math.na] 16 Nov 2017 RANDOM PERTURBATION OF LOW RANK MATRICES: IMPROVING CLASSICAL BOUNDS arxiv:3.657v5 [math.na] 6 Nov 07 SEAN O ROURKE, VAN VU, AND KE WANG Abstract. Matrix perturbation inequalities, such as Weyl s theorem

More information

Monochromatic Boxes in Colored Grids

Monochromatic Boxes in Colored Grids Monochromatic Boxes in Colored Grids Joshua Cooper, Stephen Fenner, and Semmy Purewal October 16, 008 Abstract A d-dimensional grid is a set of the form R = [a 1] [a d ] A d- dimensional box is a set of

More information

Ranks of Hadamard Matrices and Equivalence of Sylvester Hadamard and Pseudo-Noise Matrices

Ranks of Hadamard Matrices and Equivalence of Sylvester Hadamard and Pseudo-Noise Matrices Operator Theory: Advances and Applications, Vol 1, 1 13 c 27 Birkhäuser Verlag Basel/Switzerland Ranks of Hadamard Matrices and Equivalence of Sylvester Hadamard and Pseudo-Noise Matrices Tom Bella, Vadim

More information

Stochastic Design Criteria in Linear Models

Stochastic Design Criteria in Linear Models AUSTRIAN JOURNAL OF STATISTICS Volume 34 (2005), Number 2, 211 223 Stochastic Design Criteria in Linear Models Alexander Zaigraev N. Copernicus University, Toruń, Poland Abstract: Within the framework

More information

Journal of Statistical Research 2007, Vol. 41, No. 1, pp Bangladesh

Journal of Statistical Research 2007, Vol. 41, No. 1, pp Bangladesh Journal of Statistical Research 007, Vol. 4, No., pp. 5 Bangladesh ISSN 056-4 X ESTIMATION OF AUTOREGRESSIVE COEFFICIENT IN AN ARMA(, ) MODEL WITH VAGUE INFORMATION ON THE MA COMPONENT M. Ould Haye School

More information

ECE Lecture #9 Part 2 Overview

ECE Lecture #9 Part 2 Overview ECE 450 - Lecture #9 Part Overview Bivariate Moments Mean or Expected Value of Z = g(x, Y) Correlation and Covariance of RV s Functions of RV s: Z = g(x, Y); finding f Z (z) Method : First find F(z), by

More information

Dominating a family of graphs with small connected subgraphs

Dominating a family of graphs with small connected subgraphs Dominating a family of graphs with small connected subgraphs Yair Caro Raphael Yuster Abstract Let F = {G 1,..., G t } be a family of n-vertex graphs defined on the same vertex-set V, and let k be a positive

More information

L. Brown. Statistics Department, Wharton School University of Pennsylvania

L. Brown. Statistics Department, Wharton School University of Pennsylvania Non-parametric Empirical Bayes and Compound Bayes Estimation of Independent Normal Means Joint work with E. Greenshtein L. Brown Statistics Department, Wharton School University of Pennsylvania lbrown@wharton.upenn.edu

More information

Spring 2012 Math 541B Exam 1

Spring 2012 Math 541B Exam 1 Spring 2012 Math 541B Exam 1 1. A sample of size n is drawn without replacement from an urn containing N balls, m of which are red and N m are black; the balls are otherwise indistinguishable. Let X denote

More information

Refining the Central Limit Theorem Approximation via Extreme Value Theory

Refining the Central Limit Theorem Approximation via Extreme Value Theory Refining the Central Limit Theorem Approximation via Extreme Value Theory Ulrich K. Müller Economics Department Princeton University February 2018 Abstract We suggest approximating the distribution of

More information

arxiv: v1 [math.co] 20 Dec 2016

arxiv: v1 [math.co] 20 Dec 2016 F-POLYNOMIAL FORMULA FROM CONTINUED FRACTIONS MICHELLE RABIDEAU arxiv:1612.06845v1 [math.co] 20 Dec 2016 Abstract. For cluster algebras from surfaces, there is a known formula for cluster variables and

More information

Determinant of the Schrödinger Operator on a Metric Graph

Determinant of the Schrödinger Operator on a Metric Graph Contemporary Mathematics Volume 00, XXXX Determinant of the Schrödinger Operator on a Metric Graph Leonid Friedlander Abstract. In the paper, we derive a formula for computing the determinant of a Schrödinger

More information

IEOR 6711: Stochastic Models I Fall 2013, Professor Whitt Lecture Notes, Thursday, September 5 Modes of Convergence

IEOR 6711: Stochastic Models I Fall 2013, Professor Whitt Lecture Notes, Thursday, September 5 Modes of Convergence IEOR 6711: Stochastic Models I Fall 2013, Professor Whitt Lecture Notes, Thursday, September 5 Modes of Convergence 1 Overview We started by stating the two principal laws of large numbers: the strong

More information

Online Appendix. j=1. φ T (ω j ) vec (EI T (ω j ) f θ0 (ω j )). vec (EI T (ω) f θ0 (ω)) = O T β+1/2) = o(1), M 1. M T (s) exp ( isω)

Online Appendix. j=1. φ T (ω j ) vec (EI T (ω j ) f θ0 (ω j )). vec (EI T (ω) f θ0 (ω)) = O T β+1/2) = o(1), M 1. M T (s) exp ( isω) Online Appendix Proof of Lemma A.. he proof uses similar arguments as in Dunsmuir 979), but allowing for weak identification and selecting a subset of frequencies using W ω). It consists of two steps.

More information

32 Divisibility Theory in Integral Domains

32 Divisibility Theory in Integral Domains 3 Divisibility Theory in Integral Domains As we have already mentioned, the ring of integers is the prototype of integral domains. There is a divisibility relation on * : an integer b is said to be divisible

More information

Irr. Statistical Methods in Experimental Physics. 2nd Edition. Frederick James. World Scientific. CERN, Switzerland

Irr. Statistical Methods in Experimental Physics. 2nd Edition. Frederick James. World Scientific. CERN, Switzerland Frederick James CERN, Switzerland Statistical Methods in Experimental Physics 2nd Edition r i Irr 1- r ri Ibn World Scientific NEW JERSEY LONDON SINGAPORE BEIJING SHANGHAI HONG KONG TAIPEI CHENNAI CONTENTS

More information