Scribd
Upload
65 views14 pages

Feller Chap2

this is feller chapter 2

Uploaded by

Rishav Chaudhuri
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
65 views14 pages

Feller Chap2

this is feller chapter 2

Uploaded by

Rishav Chaudhuri
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
You are on page 1/ 14
An Introduction to Probability Theory and Its Applications WILLIAM FELLER Eugene Higgins Professor of Mathematics Princeton University VOLUME I SECOND EDITION John Wiley & Sons, Inc. New York . London 52 COMBINATORIAL ANALYSIS UL9 tical lines « = $ and z =n — 4. Now $ log n quite obviously ex- ceeds the area of the strip n — } Fon p> Tani ati) +1 ‘The accuracy of the approximations (9.8) is remarkable; even for n = 1 the formula leads to the two bounds 0.9958... and 1.0023... The upper bound pro- vided in (9.8) is slightly better (ef. (12.28)]. For n = 2 it yields 2.0007, for n = 5 we get 120,01..., and for n = 10 the first five significant figures are correct. PROBLEMS FOR SOLUTION Note: Sections 11 and 12 contain problems of a different character and diverse complements to the text. 10. EXERCISES AND EXAMPLES Note: Assume in each case that all arrangements have the same probability. 1. How many different sets of initials can be formed if every person has one surname and (a) exactly two given names, (}) at most two given names, (c) at most three given names? 2, In how many ways can two rooks of different colors be put on a chess- board so that they can take each other? 3, Letters in the Morse code are formed by a succession of dashes and dots with repetitions permitted. How many letters is it possible to form with ten symbols or less? 4, Each domino piece is marked by two numbers. ‘The pieces are symmetri- cal so that the number-pair is not ordered. How many different pieces can be made using the numbers 1, 2, ...,n? 5. The numbers 1, 2, ..., n are arranged in random order. Find the proba- bility that the digits (a) 1 and 2, (6) 1, 2, and 3, appear as neighbors in the order named. 6. (a) Find the probability that among three random digits there occur 2, 1, or 0 repetitions. (®) Do the same for four random digits. 7. Find the probabilities p, that in a sample of r random digits no two are equal, stimate the numerical value of pro, using Stirling's formula, ‘8. What is the probability that among k random digits (a) 0 does not appear; (6) 1 does not appear; (c) neither 0 nor 1 appears; (@) at least one of the two digits O and 1 does not appear? Let A and B represent the events in (a) and (b). Express the other events in terms of A and B. 54 COMBINATORIAL ANALYSIS (11.10 9. If n balls are placed at random into n cells, find the probability that exactly one cell remains empty. 10. At a parking lot there are twelve places arranged in a row. A man ob- served that there were eight cars parked, and that the four empty places were adjacent to each other (formed one run), Given that there are four empty places, is this arrangement surprising (indicative of non-randomness)? 11. A man is given n keys of which only one fits his door. He tries them successively (sampling without replacement). This procedure may require 1, 2, ..., n trials, Show that each of these n outcomes has probability n—!, 12. Suppose that each of n sticks is broken into one long and one short part. The 2n parts are arranged into n pairs from which new sticks are formed. Find the probability (a) that the parts will be joined in the original order, (b) that all long parts are paired with short parts." 13. Testing a statistical hypothesis. A Cornell professor got a ticket twelve times for illegal overnight parking. All twelve tickets were given either Tuesdays or Thursdays, Find the probability of this event. (Was his renting a garage only for Tuesdays and Thursdays justified?) 14. Continuation. Of twelve police tickets none was given on Sunday. Is this evidence that no tickets are given on Sundays? 15. A box contains ninety good and ten defective screws. If ten screws are used, what is the probability that none is defective? 16. From the population of five symbols a, b, c, d, ¢, a sample of size 25 is taken. Find the probability that the sample will contain five symbols of each kind. Check the result in tables of random numbers,” identifying the digits O and 1 with a, the digits 2 and 3 with b, ete. 17. Ifmmen, among whom are A and B, stand in a row, what is the probabil- ity that there will be exactly r men between A and B? If they stand in a ring instead of in a row, show that the probability is independent of r and hence 1/(m — 1), (In the circular arrangement consider only the arc leading from A to Bin the positive direction.) 18. What is the probability that two throws with three dice each will show the same configuration if (a) the dice are distinguishable, (6) they are not? 19. Show that it is more probable to get at least one ace with four dice than at least one double ace in 24 throws of two dice. (The answer is known as de Méré’s paradox. Chevalier de Méré, a gambler, thought that the two probabilities ought to be equal and blamed mathematics for his losses.) 20. From a population of n elements a sample of size r is taken, Find the probability that none of N prescribed elements will be included in the sample, 1 When cells are exposed to harmful radiation, some chromosomes break and play the role of our “sticks.” ‘The “long” side is the one containing the so-called centromere. If two “long” or two “short” parts unite, the cell dies. See D. G. Catcheside, The effect of X-ray dosage upon the frequency of induced structural changes in the chromosomes of Drosophila Melanogaster, Journal of Genetics, vol. 36 (1938), pp. 307-320. ‘ They are oceasionally extraor rily obliging: see J. A. Greenwood and E. E. Stuart, Review of Dr. Feller’s critique, Journal for Parapsychology, vol. 4 (1940), pp. 298-319, in particular p. 306. TL.10] EXERCISES AND EXAMPLES 55 assuming the sampling to be (a) without, (6) with replacement. Compare the numerical values for the two methods when (i) n = 100, r = N = 3, and (ii) n = 100, r = N = 10. 21. Spread of rumors. In a town of n+ 1 inhabitants, a person tells a rumor to a second person, who in turn repeats it to a third person, ete. At each step the recipient of the rumor is chosen at random from the n people available. Find the probability that the rumor will be told r times without: (a) returning to the originator, (b) being repeated to any person. Do the same problem when at each step the rumor is told by one person to a gathering of N randomly chosen people. (The first question is the special case N = 1.) 22. Chain letters. In a population of n + 1 people a man, the “progenitor,” sends out letters to two persons, the “first generation.” ‘These repeat the per- formance and, generally, each member of the rth generation sends out letters to two persons chosen at random. Find the probability that the generations number 1, 2, r will not include the progenitor. Find the median of the distribution, supposing n to be large. 23. A familiar problem. In a certain family four girls take turns at washing dishes. Out of a total of four breakages, three were caused by the youngest girl, and she was thereafter called clumsy. Was she justified in attributing the frequency of her breakages to chance? Discuss the connection with ran- dom placements of balls. 24, What is the probability that (a) the birthdays of twelve people will fall in twelve different calendar months (assume equal probabilities for the twelve months), (b) the birthdays of six people will fall in exactly two calendar months? 25. Given thirty people, find the probability that among the twelve months there are six containing two birthdays and six containing three. 26. A closet contains n pairs of shoes. If 2r shoes are chosen at random (with 2r ()(e-2) SY“ © so that r/n — p, then up — p® (cf. problem 13). Note: Problems 5-13 refer to the classical occupancy problem (Maxwell-Boltemann statistics): That is, r balls are distributed among n cells and each of the n" possible dis tributions has probability nT. 4 Problems 5-19 play a role in quantum statistics, the theory of photographic plates, G-M counters, etc. The formulas are therefore frequently discussed and discovered in the physical literature, usually without a realization of their classionl and essentially elementary character. Probably all the problems occur (although in modified form) in the book by Whitworth quoted at the opening of this chapter: 58 COMBINATORIAL ANALYSIS (hat 5. The probability p, that a given cell contains exactly k balls is given by the binomial distribution (4.5). The most probable number is the integer » such that (r — n + 1)/n Dept >. +> Pr} cf. problem 15.) 6. Limiting form. If n > and r — » so that the average number d = r/n of balls per cell remains constant, then (11.4) m > eX /k!. This is the Poisson distribution, discussed in chapter VI; see problem 16, 7. Let A(r,n) be the number of distributions leaving none of the n cells emply. Show by a combinatorial argument that (1.5) AG, m1) = & (1) Ak, a Conclude that (11.6) Atm) = Y(-0" (") @—»y. Hint: Use induction; assume (11.6) to hold and express A(r—k,n) in (11.5) accordingly. Change the order of summation and use the binomial formula to express A(r, n-+1) as the difference of two simple sums. Replace in the second sum » + 1 by a new index of summation and use (8.6). Note: Formula (11.6) provides a theoretical solution to an old problem but obviously it would be a thankless task to use it for the calculation of the probability x, say, that in a village of r = 1900 people every day of the year is a birthday. In chapler IV, section 2, we shail derive (11.6) by another method and obtain a simple approximation formula (showing, e.g., that x = 0.135, approximately). 8, Show that the number of distributions leaving exactly m cells empty is (11.7) En(r,n) = () A(r,n—m) = (7 zy (Cy ")a-m-m. 9. Show without using the preceding results that the probability Pmt, n) = n-TE,(r, 2) of finding exactly m cells empty satisfies am n (11.8) Pnlt +1, n) = Pm(r, n) + Pm-ate, 2) 10. Using the results of problems 7 and 8, show by direct calculation that (118) holds. Show that this method provides a new derivation (by induction on) of (11.6). 11. From (11.6) and problem 8 conclude that the probability of finding m or more cells emply és G9 Sal aah (For m = n this expression reduces to zero, as is proper.) m1) THEORETICAL PROBLEMS 59 12. The probability that each of N given cells is occupied is (11.10) u(r, n) =n = ({) 4G aya = wy* Conclude that oa me Bew(O-2 (Use the binomial theorem. For N =n we have w(r,n) = n-" A(r, n). Note that (11.11) is the analogue of (11.3) for sampling with replacement. For an alternative derivation see problem IV, 8.) 13. Limiting form. For the passage to the limit described in problem 4 one has u(r, n) > (1 — e-»)%, Note: In problems 14-19 r and n have the same meaning as above, but we assume that the balls are indistinguishable and that all distinguishable arrangements have equal probabilities (Bose-Einstein statistics). 14. The probability that a given cell contains exactly k balls is (11.12) re (Goa ye (ikea) 2 15. Show that when n > 2 zero is the most probable number of balls in any specified cell, or more precisely, go > g1 >... (cf. problem 5). 16. Limit theorem, Let n —> co and r —> ©, so that the average number of particles per cell, r/n, tends tod. ‘Then x 7 OEM (The right side is known as the geometric distribution.) 17. The probability that exactly m cells remain empty is (11.14) Pam) (oem) (uae 2): ™ Note that 1(r, n) may be interpreted as the probability that the waiting time up to the moment when the Nth element joins the sample is less than r, The result may be applied to random sampling digits: here wr, 10) — u(r — 1, 10) is the probability that a sequence of r elements must be observed to include the complete set of all ten digits. This can be used as a test of randomness, R. U, senwood (Coupon collector's test for random digits, Mathematical Tables and Other Aids to Computation, vol. 9 (1955), pp. 1-5) has tabulated the distribution and compared it to actual counts for the corresponding waiting times for the fi decimals of x and the first 2486 decimals of e, ‘The median of the waiting tir complete set of all ten digits is 27. ‘The probability that this waiting time exceeds 50 is greater than 0.05, and the probability of the waiting time exceeding 75 is about 0.0037. (11.18) 60 COMBINATORIAL ANALYSIS (1 18. The probability that a group of m prescribed cells contains a total of exactly j balls is (11.8) ox) = (MHIP Br tt eN (ates 19. Limiting form. For the passage to the limit of problem 4 we have erg tN ee. pl | m—1 )(+p4 (The right side is a special case of the negative binomial distribution to be intro- duced in chapter VI.) (11.16) qm) > ‘Theorems on Runs. In problems 20-25 we consider arrangements of r, alphas and re betas and assume that all arrangements are equally probable [see example (4.d)]. This group of problems refers to section 5a. 20. The probability that the arrangement contains exactly k runs of either kind is mail G/tais; Lyrig (uct te i Sterne of Caned when k = 2y is even, and aia) Pam {("O\CID+ClYCT Ne CE) when k = 2v + 1 is odd. 21. Continuation. Conclude that the most probable number of runs is an 2rare 2nr2 . . e hi 3. : ch integer k such that TEA nave + (Hint: Consider the ratios Pee Dom renee 22. ‘The probability that the arrangement starts with an alpha run of length » > Ois (r)sr2 + (m1 + r2)r41. (Hint: Choose the » alphas and the beta which must follow it.) What does the theorem imply for » = 0? 23. The probability of having exactly k runs of alphas is (11.19) na (1D i (C3 +1 ne ah "). Hint: This follows easily from the second part of the lemma of section 5. Alternatively, equation (11.19) may be derived from (11.17) and (11.18), but this procedure is more laborious. 24. The probability that the nth alpha is preceded by exactly m betas is actitlomea vice m+n—\y . (nt cit RENO ARS Se P"Y 25. The eae for the alphas to be arranged in k runs of which kr are of length 1, kz of i 2, ..., hy of length v (with ki -+ ...+ hy = A) is Se AEE why! (FB : y+ ny Bf B) 1.12] PROBLEMS ON BINOMIAL COEFFICIENTS 61 12, PROBLEMS AND IDENTITIES INVOLVING BINOMIAL COEFFICIENTS 1=(7)+(Q)- +. (1) +2(@) +8(5) +-- We cok ()-20)+8)-+.-0 21(5 ) +3:2(5 +4: a({) +. <= nln — )2"-2 (Hint: Use the binomial formula.) 2. Prove that for positive integers n, k 9 QQ-OG=2)+OG29-OCa-« More generally 1. For integral n > 2 (12.1) (12a) TYG e= (Gato 3. For any a>0 ao GacwC*t). If a is an integer, this can be proved also by differentiation of the geometric series Za* = (1 — 2). 4. Prove that (12.5) @) rt = (=I) (;}). 5. For integral non-negative n and r and all real @ a *(a—v) _(atl)_ (a—ny. 12) PCE NaC) (Hint: Use equation (8.6). ‘The special case n = a is frequently used.) 6. For arbitrary a and integral n > 0 (12.7) ser) =(-)" Gao) (Hint: Use equation (8.6),] ® The reader is reminded of the convention (8.5): if » runs through all integers, only finitely many terms in the sum in (12.8) are different from zero. 62 COMBINATORIAL ANALYSIS (11.12 7. For positive integers r, k a EGare tt?) (a) Prove this using (8.6). (6) Show that (12.8) is a special case of (12.7). (c) Show by an inductive argument that (12.8) leads to a new proof of the first part of the lemma of section 5. 8. In section 6 we remarked that the terms of the hypergeometric distribu- tion should add to unity. ‘This amounts to saying that for any positive integers a,b, n, a MO+QG2)+-+O-C2) Prove this by induction. (Hint: Prove first that equation (12.9) holds for a = land all.) 9. Continuation. By a comparison of the coefficients of ¢" on both sides of (12.10) A+o +0 =(14+99% prove more generally that (12.9) is true for arbitrary numbers a, b (and in- tegral n). 10. Using equation (12.9), prove that mn Q4O+O+-+ =O) 11, Using equation (12.11), prove that (12.12) 5 EAE) (?): aon —P An 12. Prove that for integers 0 0 2 N nm —N > a ee: - = 1, cae Bo ) oe G - nyt (Note that the right-hand member vanishes when r < N and when r > n,) Verify (12.18) by considering the rth derivative of ¢"-¥(t — 1)¥ at t = 1, 18. Prove by induction (using the binomial theorem) can ()E-Qpttcom (berthed (b) Next prove . (c) Generalize a= Verify (12.19) by integrating the entity 2 (1 — "= {1 — (1 — ge, 19. Show that for any positive integer m 4 mi (12.20) (@t+tytz2) =lat where the summation extends over all non-negative integers a, b, ¢, such that atb+c=m. 20. Using Stirling’s formula, prove that (12.21) (je) ~ (an) 12%, 21, Prove that for any positive integers a and b (a+ Ia +2)---@ shin) Ory (6+ 1)G+2)---6+n) al * 22. The gamma function is defined by (12.22) (12.23) T@) * da where x >0, Show that I(x) ~ (2n)e-*z*-4, (Notice that if 2 =n is an integer, Tn) = (x — 1!) 64 COMBINATORIAL ANALYSIS (1.12 23. Let a and r be arbitrary positive numbers and n a positive integer. Show that (12.24) a(a + r)(a + 2r)++-(a + mr) ~ Crm tartan Hen, [The constant C is equal to (2m)!/T(a/r).] 24, Using the results of the preceding problem, show that ala + rY(a + 2r)---@+ nr) T/r) Wb r)(b + 2r)---( + mr) ~ T(a/r) 25. Prove the following alternative form of Stirling's formula: (12.26) nl ~ (2n)n + 4)" He-in+, 26. Continuation. Using the method of the text, show that (12.27) (2m) Kn + §)"He-GD—UM4D

You might also like