Order Statistics and Benford's Law

Order Statistics and Benford's Law

Hindawi Publishing Corporation International Journal of Mathematics and Mathematical Sciences Volume 2008, Article ID 382948, 19 pages doi:10.1155/200...

2MB Sizes 0 Downloads 8 Views

Hindawi Publishing Corporation International Journal of Mathematics and Mathematical Sciences Volume 2008, Article ID 382948, 19 pages doi:10.1155/2008/382948

Research Article Order Statistics and Benford’s Law Steven J. Miller1 and Mark J. Nigrini2 1 2

Department of Mathematics and Statistics, Williams College, Williamstown, MA 01267, USA Accounting and Information Systems, School of Business, The College of New Jersey, Ewing, NJ 08628, USA

Correspondence should be addressed to Steven J. Miller, [email protected] Received 2 June 2008; Revised 6 September 2008; Accepted 13 October 2008 Recommended by Jewgeni Dshalalow Fix a base B > 1 and let ζ have the standard exponential distribution; the distribution of digits of ζ base B is known to be very close to Benford’s law. If there exists a C such that the distribution of digits of C times the elements of some set is the same as that of ζ, we say that set exhibits shifted exponential behavior base B. Let X1 , . . . , XN be i.i.d.r.v. If the Xi ’s are Unif, then as N → ∞ the distribution of the digits of the differences between adjacent order statistics converges to shifted exponential behavior. If instead Xi ’s come from a compactly supported distribution with uniformly bounded first and second derivatives and a second-order Taylor series expansion at each point, then the distribution of digits of any N δ consecutive differences and all N − 1 normalized differences of the order statistics exhibit shifted exponential behavior. We derive conditions on the probability density which determine whether or not the distribution of the digits of all the unnormalized differences converges to Benford’s law, shifted exponential behavior, or oscillates between the two, and show that the Pareto distribution leads to oscillating behavior. Copyright q 2008 S. J. Miller and M. J. Nigrini. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

1. Introduction Benford’s law gives the expected frequencies of the digits in many tabulated data. It was first observed by Newcomb in the 1880s, who noticed that pages of numbers starting with a 1 in logarithm tables were significantly more worn than those starting with a 9. In 1938 Benford 1 observed the same digit bias in a variety of phenomena. From his observations, he postulated that in many datasets, more numbers began with a 1 than with a 9; his investigations with 20,229 observations supported his belief. See 2, 3 for a description and history, and 4 for an extensive bibliography. For any base B > 1, we may uniquely write a positive x ∈ R as x  MB x · Bk , where k ∈ Z and MB x called the mantissa is in 1, B. A sequence of positive numbers {an } is Benford base B if the probability of observing a mantissa of an base B of at most s is logB s.

2

International Journal of Mathematics and Mathematical Sciences

More precisely, for s ∈ 1, B, we have

lim

N →∞

#{n ≤ N : 1 ≤ MB αn  ≤ s}  logB s. N

1.1

Benford behavior for continuous functions is defined analogously. If the functions are not positive, we study the distribution of the digits of the absolute value of the function. Thus, working base 10 we find the probability of observing a first the probability of observing a first digit of d is log10 d1−log10 d, implying that about 30% of the time the first digit is a 1. We can prove many mathematical systems follow Benford’s law, ranging from recurrence relations 5 to n! 6 to iterates of power, exponential and rational maps, as well as Newton’s method 7–9; to chains of random variables and hierarchical Bayesian models 10; to values of L-functions near the critical line; to characteristic polynomials of random matrix ensembles and iterates of the 3x  1-Map 11, 12; as well as to products of random variables 13. We also see Benford’s law in a variety of natural systems, such as atomic physics 14, biology 15, and geology 16. Applications of Benford’s law range from rounding errors in computer calculations see 17, page 255 to detecting tax see 18, 19 and voter fraud see 20. This work is motivated by two observations see Remark 1.9 for more details. First, since Benford’s seminal paper, many investigations have shown that amalgamating data from different sources leads to Benford behavior; second, many standard probability distributions are close to Benford behavior. We investigate the distribution of digits of differences of adjacent ordered random variables. For any δ < 1, if we study at most N δ consecutive differences of a dataset of size N, the resulting distribution of leading digits depends very weakly on the underlying distribution of the data, and closely approximates Benford’s law. We then investigate whether or not studying all the differences leads to Benford behavior; this question is inspired by the first observation above, and has led to new tests for data integrity see 21. These tests are quick and easy-to-apply, and have successfully detected problems with some datasets, thus providing a practical application of our main results. Proving our results requires analyzing the distribution of digits of independent random variables drawn from the standard exponential, and quantifying how close the distribution of digits of a random variable with the standard exponential distribution is to Benford’s law. Leemis et al. 22 have observed that the standard exponential is quite close to Benford’s law; this was proved by Engel and Leuenberger 23, who showed that the maximum difference in the cumulative distribution function from Benford’s law base 10 is at least .029 and at most .03. We provide an alternate proof of this result in the appendix using a different technique, as well as showing that there is no base B such that the standard exponential distribution is Benford base B Corollary A.2. Both proofs apply Fourier analysis to periodic functions. In 23, equation 5, the main step is interchanging an integration and a limit. Our proof is based on applying Poisson summation to the derivative of the cumulative distribution function of the logarithms modulo 1, FB . Benford’s law is equivalent to FB b  b, which by calculus is the same as FB b  1 and FB 0  0. Thus, studying the deviation of FB b from 1 is a natural way to investigate the deviations from Benford behavior. We hope the details of these calculations may be of use to others in investigating related problems Poisson summation has been fruitfully used by Kontorovich and Miller 11 and Jang et al. 10 in proving many systems are Benford’s; see also 24.

S. J. Miller and M. J. Nigrini

3

1.1. Definitions A sequence {an }∞ n1 ⊂ 0, 1 is equidistributed if lim

N →∞

#{n : n ≤ N, an ∈ a, b} b−a N

∀a, b ⊂ 0, 1.

1.2

Similarly, a continuous random variable on 0, ∞, whose probability density function is p, is equidistributed modulo 1 if T lim

T →∞

0

χa,b xpxdx  b − a, T pxdx 0

1.3

for any a, b ⊂ 0, 1, where χa,b x  1 for x mod 1 ∈ a, b and 0 otherwise. A positive sequence or values of a function is Benford base B if and only if its base B logarithms are equidistributed modulo 1; this equivalence is at the heart of many investigations of Benford’s law see 6, 25 for a proof. We use the following notations for the various error terms. 1 Let Ex denote an error of at most x in absolute value; thus fb  gb  Ex means |fb − gb| ≤ x. 2 Big-Oh notation: for gx a nonnegative function, we say fx  Ogx if there exist an x0 and a C > 0 such that, for all x ≥ x0 , |fx| ≤ Cgx. The following theorem is the starting point for investigating the distribution of digits of order statistics. Theorem 1.1. Let ζ have the standard (unit) exponential distribution   Prob ζ ∈ α, β 



e−t dt,

α, β ∈ 0, ∞.

1.4

α

For b ∈ 0, 1, let FB b be the cumulative distribution function of logB ζ mod 1; thus FB b : ProblogB ζ mod 1 ∈ 0, b. Then, for all M ≥ 2, FB b

   2πim Re e−2πimb Γ 1  log B m1

12

∞ 

  

√ 2πim 2 Re e−2πimb Γ 1   E 4 2πc1 Be−π −c2 BM/ log B , log B m1

M−1 

12

1.5

4

International Journal of Mathematics and Mathematical Sciences

where c1 B, c2 B are constants such that for all m ≥ M ≥ 2, one has 2

e

2π 2 m/ log B

−e

−2π 2 m/ log B

e2π m/ log B ≥ , c12 B

m ≤ e2c2 Bm/ log B , log B 1 2 1 − eπ −c2 BM/ log B ≥ √ . 2 For B ∈ e, 10, take c1 B 



1.6

2 and c2 B  1/5, which give

  Prob log ζ mod 1 ∈ a, b       2r b−a · sin πb  a  θ · sin πb − a  E 6.32 · 10−7 , π

1.7

with r ≈ 0.000324986, θ ≈ 1.32427186, and       2r1 Prob log10 ζ mod 1 ∈ a, b  b − a  sin πb  a − θ1 · sin πb − a π 1.8       r2 − sin 2πb  a  θ2 · sin 2πb − a  E 8.5 · 10−5 π with r1 ≈ 0.0569573,

θ1 ≈ 0.8055888,

r2 ≈ 0.0011080,

θ2 ≈ 0.1384410.

1.9

The above theorem was proved in 23; we provide an alternate proof in Appendix A. As remarked earlier, our technique consists of applying Poisson summation to the derivative of the cumulative distribution function of the logarithms modulo 1; it is then very natural and easy to compare deviations from the resulting distribution and the uniform distribution if a dataset satisfies Benford’s law, then the distribution of its logarithms is uniform. Our series expansions are obtained by applying properties of the Gamma function. Definition 1.2 Definition exponential behavior, shifted exponential behavior. Let ζ have the standard exponential distribution, and fix a base B. If the distribution of the digits of a set is the same as the distribution of the digits of ζ, then one says that the set exhibits exponential behavior base B. If there is a constant C > 0 such that the distribution of digits of all elements multiplied by C is exponential behavior, then one says that the system exhibits shifted exponential behavior with shift of logB C mod 1. We briefly describe the reasons behind this notation. One important property of Benford’s law is that it is invariant under rescaling; many authors have used this property to characterize Benford behavior. Thus, if a dataset is Benford base B, and we fix a positive number C, so is the dataset obtained by multiplying each element by C. This is clear if, instead of looking at the distribution of the digits, we study the distribution of the base B

S. J. Miller and M. J. Nigrini

5

logarithms modulo 1. Benford’s law is equivalent to the logarithms modulo 1 being uniformly distributed see, e.g., 6, 25; the effect of multiplying all entries by a fixed constant simply translates the uniform distribution modulo 1, which is again the uniform distribution. The situation is different for exponential behavior. Multiplying all elements by a fixed constant C where C  / Bk for some k ∈ Z does not preserve exponential behavior; however, the effect is easy-to-describe. Again looking at the logarithms, exponential behavior is equivalent to the base B logarithms modulo 1 having a specific distribution which is almost equal to the uniform distribution at least if the base B is not too large. Multiplying by a fixed constant C /  B k shifts the logarithm distribution by logB C mod 1. 1.2. Results for differences of orders statistics We consider a simple case first, and show how the more general case follows. Let X1 , . . . , XN be independent identically distributed from the uniform distribution on 0, L. We consider L fixed and study the limit as N → ∞. Let X1:N , . . . , XN:N be the Xi ’s in increasing order. The Xi:N are called the order statistics, and satisfy 0 ≤ X1:N ≤ X2:N ≤ · · · ≤ XN:N ≤ L. We investigate the distribution of the leading digits of the differences between adjacent Xi:N ’s, Xi1:N − Xi:N . For convenience, we periodically continue the data and set XiN:N  Xi:N  L. As we have N differences in an interval of size L, the average value of Xi1:N − Xi:N is of size L/N, and it is sometimes easier to study the normalized differences Zi;N 

Xi1:N − Xi:N . L/N

1.10

As the Xi ’s are drawn from a uniform distribution, it is a standard result that as N → ∞, the Zi;N ’s are independent random variables, each having the standard exponential distribution. b Thus, as N → ∞, the probability that Zi;N ∈ a, b tends to a e−t dt see 26, 27 for proofs. For uniformly distributed random variables, if we know the distribution of logB Zi;N mod 1, then we can immediately determine the distribution of the digits of the Xi1:N − Xi:N base B because logB Zi;N  logB

Xi1:N − Xi:N L/N







 logB Xi1:N − Xi:N − logB



 L . N

1.11

As Zi;N are independent with the standard exponential distribution as N → ∞; if Xi are independent uniformly distributed, the behavior of the digits of the differences Xi1:N − Xi:N is an immediate consequence of Theorem 1.1. Theorem 1.3 Shifted exponential behavior of differences of independent uniformly distributed random variables. Let X1 , . . . , XN be independently distributed from the uniform distribution on 0, L, and let X1:N , . . . , XN:N be Xi ’s in an increasing order. As N → ∞, the distribution of the digits (base B) of the differences Xi1:N − Xi:N converges to shifted exponential behavior, with a shift of logB L/N mod 1. A similar result holds for other distributions.

6

International Journal of Mathematics and Mathematical Sciences

Theorem 1.4 Shifted exponential behavior of subsets of differences of independent random variables. Let X1 , . . . , XN be independent, identically distributed random variables whose density fx has a second-order Taylor series at each point with first and second derivatives uniformly bounded, and let the Xi:N ’s be the Xi ’s in increasing order. Fix a δ ∈ 0, 1. Then, as N → ∞ the distribution of the digits (base B) of N δ consecutive differences Xi1:N − Xi:N converges to shifted exponential behavior, provided that Xi:N ’s are from a region where fx is nonzero. The key ingredient in this generalization is that the techniques, which show that the differences between uniformly distributed random variables become independent exponentially distributed random variables, can be modified to handle more general distributions. We restricted ourselves to a subset of all consecutive spacings because the normalization factor changes throughout the domain. The shift in the shifted exponential behavior depends on which set of N δ differences we study, coming from the variations in the normalizing factors. Within a bin of N δ differences, the normalization factor is basically constant, and we may approximate our density with a uniform distribution. It is possible for these variations to cancel and yield Benford behavior for the digits of all the unnormalized differences. Such a result is consistent with the belief that amalgamation of data from many different distributions becomes Benford; however, this is not always the case see Remark 1.6. From Theorems 1.1 and 1.4, we obtain the following theorem. Theorem 1.5 Benford behavior for all the differences of independent random variables. Let X1 , . . . , XN be independent identically distributed random variables whose density fx is compactly supported and has a second-order Taylor series at each point with first and second derivatives uniformly bounded. Let the Xi:N ’s be the Xi ’s in an increasing order Fx be the cumulative distribution function for fx, and fix a δ ∈ 0, 1. Let I , δ, N   N 1−δ , N 1−δ − N 1−δ . For each fixed ∈ 0, 1/2, assume that i fF −1 kN δ−1  is not too small for k ∈ I , δ, N lim

N →∞

minN − δ/2 , N δ−1   0, k∈I ,δ,N fF −1 kN δ−1 

1.12

max

ii logB fF −1 kN δ−1  mod 1 is equidistributed: for all α, β ⊂ 0, 1 lim

N →∞

#{k ∈ I , δ, N : logB fF −1 kN δ−1  mod 1 ∈ α, β} Nδ

 β − α.

1.13

Then, if > max0, 1/3 − δ/2 and < δ/2, the distribution of the digits of the N − 1 differences Xi1:N − Xi:N converges to Benford’s law (base B) as N → ∞. Remark 1.6. The conditions of Theorem 1.5 are usually not satisfied. We are unaware of any situation where 1.13 holds; we have included Theorem 1.5 to give a sufficient condition of what is required to have Benford’s law satisfied exactly, and not just approximately. In Lemma 3.3, we show with: Example 3.3 shows that the conditions fail for the Pareto distribution, and the limiting behavior oscillates between Benford and a sum of shifted exponential behavior. If several datasets each exhibit shifted exponential behavior but with distinct shifts, then

S. J. Miller and M. J. Nigrini

7 0.03

0.2

0.025

0.175

0.02

0.15

0.015

0.125 0.075

0.01 2

4

0.05

6

8

10

0.005 −0.005

a

0.2

0.4

0.6

0.8

1

b

Figure 1: All 499 999 differences of adjacent order statistics from 500 000 independent random variables from the Pareto distribution with minimum value and variance 1. a Observed digits of scaled differences of adjacent random variables versus Benford’s law; b scaled observed minus Benford’s law cumulative distribution of base 10 logarithms.

the amalgamated dataset is closer to Benford’s law than any of the original datasets. This is apparent by studying the logarithms modulo 1. The differences between these densities and Benford’s law will look like Figure 1b except, of course, that different shifts will result in shifting the plot modulo 1. The key observation is that the unequal shifts mean that we do not have reinforcements from the peaks of modulo 1 densities being aligned, and thus the amalgamation will decrease the maximum deviations. The arguments generalize to many densities whose cumulative distribution functions have tractable closed-form expressions x e.g., exponential, Weibull, or fx  e−e ex . The situation is very different if instead we study normalized differences:

i:N  Xi1:N − Xi:N , Z 1/NfXi:N 

1.14

note if fx  1/L is the uniform distribution on 0, L, 1.14 reduces to 1.10. Theorem 1.7 Shifted exponential behavior for all the normalized differences of independent random variables. Assume the probability distribution f satisfies the conditions of Theorem 1.5 and

i;N is as in 1.14. Then, as N → ∞, the distribution of the digits of Z

i:N converges to 1.12 and Z shifted exponential behavior. Remark 1.8. Appropriately scaled, the distribution of the digits of the differences is universal, and is the exponential behavior of Theorem 1.1. Thus, Theorem 1.7 implies that the natural quantity to study is the normalized differences of the order statistics, not the differences see also Remark 3.5. With additional work, we could study densities with unbounded support and show that, through truncation, we can get arbitrarily close to shifted exponential behavior. Remark 1.9. The main motivation for this work is the need for improved ways of assessing the authenticity and integrity of scientific and corporate data. Benford’s law has been successfully applied to detecting income tax, corporate, and voter fraud see 18–20; in 21, we use these results to derive new statistical tests to examine data authenticity and integrity. Early applications of these tests to financial data showed that it could detect errors in data downloads, rounded data, and inaccurate ordering of data. These attributes are not

8

International Journal of Mathematics and Mathematical Sciences

easily observable from an analysis of descriptive statistics, and detecting these errors can help managers avoid costly decisions based on erroneous data. The paper is organized as follows. We prove Theorem 1.1 in Appendix A by using Poisson summation to analyze FB b. Theorem 1.3 follows from the results of the order statistics of independent uniform variables. The proof of Theorem 1.4 is similar, and given in Section 2. In Section 3, we prove Theorems 1.5 and 1.7. 2. Proofs of Theorems 1.3 and 1.4 Theorem 1.3 is a consequence of the fact that the normalized differences between the order statistics drawn from the uniform distribution converge to being independent standard exponentials. The proof of Theorem 1.4 proceeds similarly. Specifically, over a short enough region, any distribution with a second-order Taylor series at each point with first and second derivatives uniformly bounded is well approximated by a uniform distribution. To prove Theorem 1.4, it suffices to show that if X1 , . . . , XN are drawn from a sufficiently nice distribution, then for any fixed δ ∈ 0, 1 the limiting behavior of the order statistics of N δ adjacent Xi ’s becomes Poissonian i.e., the N δ − 1 normalized differences converge to being independently distributed from the standard exponential. We prove this below for compactly supported distributions fx that have a second-order Taylor series at each point with the first and second derivatives uniformly bounded, and when the N δ adjacent Xi ’s are from a region where fx is bounded away from zero. b For each N, consider intervals aN , bN  such that aNN fxdx  N δ /N; thus, the proportion of the total mass in such intervals is N δ−1 . We fix such an interval for our arguments. For each i ∈ {1, . . . , N}, let  1, wi  0,

  if Xi ∈ aN , bN otherwise.

2.1

Note wi is 1 with probability N δ−1 and 0 with probability 1 − N δ−1 ; wi is a binary indicator random variable, telling us whether or not Xi ∈ aN , bN . Thus,   N  E wi  N δ , i1



N  Var wi



   N δ · 1 − N δ−1 .

2.2

i1

Let MN be the number of Xi in aN , bN , and let βN be any nondecreasing sequence tending to infinity in the course of the proof, we will find that we may take any sequence with βN  ON δ/2 . By 2.2 and the central limit theorem which we may use as wi ’s satisfy the Lyapunov condition, with probability tending to 1, we have   MN  N δ  O βN N δ/2 .

2.3

We assume that in the interval aN , bN  there exist constants c and C such that whenever x ∈ aN , bN , 0 < c < fx < C < ∞; we assume that these constants hold for all regions investigated and for all N. If our distribution has unbounded support, for any

S. J. Miller and M. J. Nigrini

9

> 0, we can truncate it on both sides so that the omitted probability is at most . Our result is then trivially modified to be within of shifted exponential behavior. Thus,   c · bN − aN ≤

 bN

  fxdx  N δ−1 ≤ C bN − aN ,

2.4

aN

implying that bN − aN is of size N δ−1 . If we assume that fx has at least a second-order Taylor expansion, then       2  fx  f aN  f  aN x − aN  O x − aN         f aN  f  aN x − aN  O N 2δ−2 .

2.5

As we assume that the first and second derivatives are uniformly bounded, as well as f being bounded away from zero in the intervals under consideration, all Big-Oh constants below are independent of N. Thus, bN − aN 

  N δ−1  O N 2δ−2 . faN 

2.6

We now investigate the order statistics of the MN of the Xi ’s that lie in aN , bN . We b know aNN fxdx  N δ−1 ; by setting gN x  fxN 1−δ , then gN x is the conditional density function for Xi , given that Xi ∈ aN , bN . Thus, gN x integrates to 1, and for x ∈ aN , bN , we have        gN x  f aN · N 1−δ  f  aN x − aN · N 1−δ  O N δ−1 .

2.7

We have an interval of size N δ−1 /faN   ON 2δ−2 , and MN  N δ  OβN N δ/2  of the Xi lying in the interval remember that βN are any nondecreasing sequence tending to infinity. Thus, with probability tending to 1, the average spacing between adjacent ordered Xi is   N δ−1 /faN   ON 2δ−2     −1  f aN N  N −1 · O βN N −δ/2  N δ−1 , MN

2.8

in particular, we see that we must choose βN  ON δ/2 . As δ ∈ 0, 1, if we fix a k such that Xk ∈ aN , bN , then we expect the next Xi to the right of Xk to be about t/NfaN  units away, where t is of size 1. For a given Xk , we can compute the conditional probability that the next Xi is between t/NfaN  and t  Δt/NfaN  units to the right. It is simply the difference of the probability that all the other MN − 1 of the Xi ’s in aN , bN  are not in the interval Xk , Xk  t/NfaN  and the probability that all other Xi in aN , bN  are not in the interval Xk , Xk  t  Δt/NfaN ; note that we are using the wrapped interval aN , bN . Some care is required in these calculations. We have a conditional probability as we assume that Xk ∈ aN , bN  and that exactly MN of the Xi are in aN , bN . Thus, these

10

International Journal of Mathematics and Mathematical Sciences

probabilities depend on two random variables, namely, Xk and MN . This is not a problem in practice, however e.g., MN is tightly concentrated about its mean value. Recalling our expansion for gN x and that bN − aN  N δ−1 /faN   ON 2δ−2  and t is of size 1, after simple algebra, we find that with probability tending to 1, for a given Xk and MN , the first probability is  Xk t/NfaN   MN −1 1− gN xdx .

2.9

Xk

The above integral equals tN δ  ON −1  use the Taylor series expansion in 2.7 and note that the interval aN , bN  is of size ON δ−1 . Using 2.3, it is easy to see that this is a.s. equal to 

t  ON δ−1  βN N −δ/2  1− MN

MN −1 .

2.10

We, therefore, find that as N → ∞, the probability that MN − 1 of the Xi ’s i  / k are in aN , bN  \ Xk , Xk  t/NfaN , conditioned on Xk and MN , converges to e−t . Some care is required, as the exceptional set in our a.s. statement can depend on t. This can be surmounted by taking expectations with respect to our conditional probabilities and applying the dominated convergence theorem. The calculation of the second probability, the conditional probability that the MN − 1 other Xi ’ that are aN , bN  not in the interval Xk , Xk  t  Δt/NfaN , given Xk and MN , follows analogously by replacing t with t  Δt in the previous argument. We thus find that this probability is e−tΔt . As  tΔt

e−u du  e−t − e−tΔt ,

2.11

t

we find that the density of the difference between adjacent order statistics tends to the standard unit exponential density; thus, the proof of Theorem 1.4 now follows from Theorem 1.3.

3. Proofs of Theorems 1.5 and 1.7 We generalize the notation from Section 2. Let fx be any distribution with a second-order Taylor series at each point with first and second derivatives uniformly bounded, and let X1:N , . . . , XN:N be the order statistics. We fix a δ ∈ 0, 1, and for k ∈ {1, . . . , N 1−δ }, we consider bins ak;N , bk;N  such that  bk;N ak;N

fxdx 

Nδ  N δ−1 , N

3.1

S. J. Miller and M. J. Nigrini

11

there are N 1−δ such bins. By the central limit theorem see 2.3, if Mk;N is the number of order statistics in ak;N , bk;N , then, provided that > max0, 1/3 − δ/2 with probability tending to 1, we have   Mk;N  N δ  O N δ/2 ,

3.2

of course we also require < δ/2, as, otherwise, the error term is larger than the main term. Remark 3.1. Before we considered just one fixed interval; as we are studying N 1−δ intervals simultaneously, we need in the exponent so that with high probability, all intervals have to first order N δ order statistics. For the arguments below, it would have sufficed to have an error of size ON δ− . We thank the referee for pointing out that > 1/3 − δ/2, and provide his argument in Appendix B. Similar to 2.8, the average spacing between adjacent order statistics in ak;N , bk;N  is   −1     N −1 · O N − δ/2  N δ−1 . f ak;N N

3.3

Note that 3.3 is the generalization of 1.11; if f is the uniform distribution on 0, L, then fak;N   1/L. By Theorem 1.4, as N → ∞, the distribution of digits of the differences in each bin converges to shifted exponential behavior; however, the variation in the average spacing between bins leads to bin-dependent shifts in the shifted exponential behavior. Similar to 1.11, we can study the distribution of digits of the differences of the normalized order statistics. If Xi:N and Xi1:N are in ak;N , bk;N , then Zi;N 

logB Zi;N

Xi1:N − Xi:N

, fak;N N  N −1 · ON − δ/2  N δ−1      −1   logB Xi1:N − Xi:N  logB N − logB f ak;N  ON − δ/2  N δ−1 . −1

3.4

Note we are using the same normalization factor for all differences between adjacent order statistics in a bin. Later, we show that we may replace fak;N  with fXi:N . As we study all Xi1:N − Xi:N in the bin ak;N , bk;N , it is useful to rewrite the above as      −1  logB Xi1:N − Xi:N  logB Zi;N − logB N  logB f ak;N  O N − δ/2  N δ−1 .

3.5

We have N 1−δ bins, so k ∈ {1, . . . , N 1−δ }. As we only care about the limiting behavior, we may safely ignore the first and last bins. We may, therefore, assume that each ak;N is finite, and ak1;N  bk;N . Of course, we know that both quantities are finite as we assumed that our distribution has compact support. We remove the last bins to simplify generalizations to noncompactly supported distributions. Let Fx be the cumulative distribution function for fx. Then, Fak;N  

k − 1N δ  k − 1N δ−1 . N

3.6

12

International Journal of Mathematics and Mathematical Sciences

For notational convenience, we relabel the bins so that k ∈ {0, . . . , N 1−δ − 1}; thus Fak;N   kN δ−1 . We now prove our theorems which determine when these bin-dependent shifts cancel yielding Benford behavior, or reinforce yielding sums of shifted exponential behavior. Proof of Theorem 1.5. There are approximately N δ differences in each bin ak;N , bk;N . By Theorem 1.4, the distribution of the digits of the differences in each bin converges to shifted exponential behavior. As we assume that the first and second derivatives of f are uniformly bounded, the Big-Oh constants in Section 2 are independent of the bins. The shift in the shifted exponential behavior in each bin is controlled by the last two terms on the righthand side of 3.5. The logB N shifts the shifted exponential behavior in each bin equally. The bin-dependent shift is controlled by the final term      −1  minN − δ/2 , N δ−1  logB f ak;N  O N − δ/2  N δ−1  −logB fak;N   logB 1  . fak;N  3.7 Thus, each of the N 1−δ bins exhibits shifted exponential behavior, with a bindependent shift composed of the two terms in 3.7. By 1.12, fak;N  are not small compared to minN − δ/2 , N δ−1 , and hence the second term logB 1  minN − δ/2 , N δ−1 /fak;N  is negligible. In particular, this factor depends only very weakly on the bin, and tends to zero as N → ∞. Thus, the bin-dependent shift in the shifted exponential behavior is approximately −logB fak;N   −logB fF −1 kN δ−1 . If these shifts are equidistributed modulo 1, then the deviations from Benford behavior cancel, and the shifted exponential behavior of each bin becomes Benford behavior for all the differences. Remark 3.2. Consider the case when the density is a uniform distribution on some interval. Then, all fF −1 kN δ−1  are equal, and each bin has the same shift in its shifted exponential behavior. These shifts, therefore, reinforce each other, and the distribution of all the differences is also shifted exponential behavior, with the same shift. This is observed in numerical experiments see Theorem 1.3 for an alternate proof. We analyze the assumptions of Theorem 1.5. The condition from 1.12 is easy-tocheck, and is often satisfied. For example, if the probability density is a finite union of monotonic pieces and is zero only finitely often, then 1.12 holds. This is because for k ∈ I , δ, N, F −1 kN δ−1  ∈ F −1  , F −1 1 − , and this is, therefore, independent of N if f vanishes finitely often, we need to remove small subintervals from I , δ, N, but the analysis proceeds similarly. The only difficulty is basically a probability distribution with intervals of zero probability. Thus, 1.12 is a mild assumption. If we choose any distribution other than a uniform distribution, then fx is not constant; however, 1.13 does not need to hold i.e., logB fak;N  mod 1does not need to be equidistributed as N → ∞. For example, consider a Pareto distribution with minimum value 1 and exponent a > 0. The density is  fx 

ax−a−1 0

if x ≥ 1, otherwise.

3.8

S. J. Miller and M. J. Nigrini

13

The Pareto distribution is known to be useful in modelling natural phenomena, and for appropriate choices of exponents, it yields approximately Benford behavior see 16. Example 3.3. If f is a Pareto distribution with minimum value 1 and exponent a > 0, then f does not satisfy the second condition of Theorem 1.5, 1.13. To see this, note that the cumulative distribution function of f is Fx  1 − x−a . As we only care about the limiting behavior, we need only to study k ∈ I , δ, N   N 1−δ , N 1−δ − N 1−δ . Therefore, Fak;N   kN δ−1 implies that  −1/a ak;N  1 − kN δ−1 ,

   a1/a f ak;N  a 1 − kN δ−1 .

3.9

The condition from 1.12 is satisfied, namely, lim

N →∞

minN − δ/2 , N δ−1   lim N →∞ fak;N  k∈I ,δ,N max

max

minN − δ/2 , N δ−1 

k∈I ,δ,N

akN δ−1 

a1/a

 0,

3.10

as k is of size N 1−δ . Let j  N 1−δ − k ∈ I , δ, N. Then, the bin-dependent shifts are   a1   logB 1 − kN δ−1  logB a logB f ak;N  a   a1  logB jN 1−δ  logB a a      logB j a1/a  logB aN 1−δa1/a .

3.11

Thus, for a Pareto distribution with exponent a, the distribution of all the differences becomes Benford if and only if j a1/a is Benford. This follows from the fact that a sequence is Benford if and only if its logarithms are equidistributed. For fixed m, j m is not Benford e.g., 6, and thus the condition from 1.13 fails. Remark 3.4. We chose to study a Pareto distribution because the distribution of digits of a random variable drawn from a Pareto distribution converges to Benford behavior base 10 as a → 1; however, the digits of the differences do not tend to Benford or shifted exponential behavior. A similar analysis holds for many distributions with good closed-form expressions for the cumulative distribution function. In particular, if f is the density of an exponential x or Weibull distribution or fx  e−e ex , then f does not satisfy the second condition of Theorem 1.5, 1.13. Modifying the proof of Theorem 1.5 yields our result on the distribution of digits of the normalized differences. Proof of Theorem 1.7. If f is the uniform distribution, there is nothing to prove. For general f, rescaling the differences eliminates the bin-dependent shifts. Let

i:N  Xi1:N − Xi:N . Z 1/NfXi:N 

3.12

14

International Journal of Mathematics and Mathematical Sciences

In Theorem 1.5, we use the same scale factor for all differences in a bin see 3.4. As we assume the first and second derivatives of f are uniformly bounded, 2.5 and 2.6 imply that for Xi:N ∈ ak;N , bk;N ,       f Xi:N  f ak;N  O bk;N − ak;N     N δ−1  f ak;N  O  N 2δ−2 , fak;N 

3.13

and the Big-Oh constants are independent of k. As we assume that f satisfies 1.12, the error term is negligible. Thus, our assumptions on f imply that f is basically constant on each bin, and we may replace the local rescaling factor fXi:N  with the bin rescaling factor fak;N . Thus, each bin of normalized differences has the same shift in its shifted exponential behavior. Therefore all the shifts reinforce, and the digits of all the normalized differences exhibit shifted exponential behavior as N → ∞. As an example of Theorem 1.7, in Figure 1 we consider 500,000 independent random variables drawn from the Pareto distribution with exponent a

4

  √ √ 3 3 19 − 3 33  19  3 33 3

3.14

we chose a to make the variance equal 1. We study the distribution of the digits of the differences in base 10. The amplitude is about .018, which is the amplitude of the shifted exponential behavior of Theorem 1.1 see the equation in 23, Theorem 2 or 1.5 of Theorem 1.1. Remark 3.5. The universal behavior of Theorem 1.7 suggests that if we are interested in the behavior of the digits of all the differences, the natural quantity to study is the normalized differences. For any distribution with uniformly bounded first and second derivatives and a second-order Taylor series expansion at each point, we obtain shifted exponential behavior. Appendices A. Proof of Theorem 1.1 To prove Theorem 1.1, it suffices to study the distribution of logB ζ mod 1 when ζ has the standard exponential distribution see 1.4. We have the following useful chain of equalities. Let a, b ⊂ 0, 1. Then, ∞       Prob logB ζ ∈ a  k, b  k Prob logB ζ mod 1 ∈ a, b  k−∞



∞ 

   Prob ζ ∈ Bak , Bbk

k−∞



∞  

e−B

k−∞

ak

− e−B

bk



.

A.1

S. J. Miller and M. J. Nigrini

15

It suffices to investigate A.1 in the special case when a  0, as the probability of any interval α, β can always be found by subtracting the probability of 0, α from 0, β. We are, therefore, led to study, for b ∈ 0, 1, the cumulative distribution function of logB ζ mod 1, ∞     −Bk bk  − e−B . FB b : Prob logB ζ mod 1 ∈ 0, b  e

A.2

k−∞

This series expansion converges rapidly, and Benford behavior for ζ is equivalent to the rapidly converging series in A.2 equalling b for all b. As Benford behavior is equivalent to FB b equals b for all b ∈ 0, 1, it is natural to compare FB b to 1. If the derivatives were identically 1, then FB b would equal b plus some constant. However, A.2 is zero when b  0, which implies that this constant would be zero. It is hard to analyze the infinite sum for FB b directly. By studying the derivative FB b, we u bu find a function with an easier Fourier transform than the Fourier transform of e−B − e−B , which we then analyze by applying Poisson summation. We use the fact that the derivative of the infinite sum FB b is the sum of the derivatives of the individual summands. This is justified by the rapid decay of the summands see, e.g., 28, Corollary 7.3. We find FB b 

∞ 

∞ 

e−B Bbk log B  bk

k−∞

e−βB βBk log B, k

A.3

k−∞

where for b ∈ 0, 1, we set β  Bb . t Let Ht  e−βB βBt log B; note β ≥ 1. As Ht is of rapid decay in t, we may apply Poisson summation e.g., 29. Thus, ∞ 

∞ 

Hk 

k−∞

 is the Fourier transform of H: Hu  where H  FB b



∞ 

Hk 

k−∞

∞ 

 Hk,

A.4

k−∞

 Hk 

k−∞

∞

−∞

Hte−2πitu dt. Therefore,

∞ ∞  k−∞ −∞

e−βB βBt log B · e−2πitk dt. t

A.5

Let us change variables by taking w  Bt . Thus, dw  Bt log B dt or dw/w  log B dt. As e−2πitk  Bt/ log B −2πik  w−2πik/ log B , we have FB b 

∞ ∞  k−∞



∞ 

0

β2πik/ log B

k−∞



e−βw βw · w−2πik/ log B ∞

e−u u−2πik/ log B du

0

  2πik , β2πik/ log B Γ 1 − log B k−∞ ∞ 

dw w A.6

16

International Journal of Mathematics and Mathematical Sciences

where we have used the definition of the Γ-function Γs 

∞

e−u us−1 du,

Res > 0.

A.7

    2πim 2πim −2πim/ log B . β Γ 1− Γ 1 log B log B

A.8

0

As Γ1  1, we have FB b

1

∞ 



m1

β

2πim/ log B

Remark A.1. The above series expansion is rapidly convergent, and shows the deviations of logB ζ mod 1 from being equidistributed as an infinite sum of special values of a standard function. As β  Bb , we have β2πim/ log B  cos2πmb  i sin2πmb, which gives a Fourier series expansion for F  b with coefficients arising from special values of the Γ-function. We can improve A.8 by using additional properties of the Γ-function. If y ∈ R, then from A.7, we have Γ1 − iy  Γ1  iy where the bar denotes complex conjugation. Thus, the mth summand in A.8 is the sum of a number and its complex conjugate, which is simply twice the real part. We have formulas for the absolute value of the Γ-function for large argument. We use see 30, page 946, equation 8.332 that   Γ1  ix2 

2πx πx  . sinhπx eπx − e−πx

A.9

Writing the summands in A.8 as 2Ree−2πimb Γ1  2πim/ log B, A.8 becomes FB b

      ∞  2πim 2πim −2πimb −2πimb 2 . 12 Re e Γ 1 Re e Γ 1 log B log B m1 mM M−1 

A.10

The rest of the claims of Theorem 1.1 follow from simple estimation, algebra, and trigonometry. With constants as in the theorem, if we take M  1 and B  e resp., B  10 the error is at most .00499 resp., .378, while if M  2 and B  e resp., B  10, the error is at most 3.16 · 10−7 resp., .006. Thus, just one term is enough to get approximately five digits of accuracy base e, and two terms give three digits of accuracy base 10. For many bases, we have reduced the problem to evaluate Ree−2πib Γ1  2πi/ log B. This example illustrates the power of Poisson summation, taking a slowly convergent series expansion and replacing it with a rapidly converging one. Corollary A.2. Let ζ have the standard exponential distribution. There is no base B > 1 such that ζ is Benford base B. Proof. Consider the infinite series expansion in 1.5. As e−2πimb is a sum of a cosine and a sine term, 1.5 gives a rapidly convergent Fourier series expansion. If ζ were Benford base B, then FB b must be identically 1; however, Γ1  2πim/ log B is never zero for m a positive integer because its modulus is nonzero see A.9. As there is a unique rapidly convergent

S. J. Miller and M. J. Nigrini

17

Fourier series equal to 1 namely, gb  1; see 29 for a proof, our FB b cannot identically equal 1. B. Analyzing N 1−δ intervals simultaneously We show why in addition to > 0 we also needed > 1/3 − δ/2 when we analyzed N 1−δ intervals simultaneously in 3.2; we thank one of the referees for providing this detailed argument. . , YN be i.i.d.r.v. with EYi   0, VarYi   σ 2 , E|Yi |3  < ∞, and set SN  Let Y1 , . . √ Y1  · · ·  YN / Nσ 2 . Let Φx denote the cumulative distribution function of the standard normal. Using a nonuniform sharpening of the Berry-Ess´een estimate e.g., 31, we find that for some constant c > 0,     Prob SN ≤ x − Φx ≤

cE|Y1 |3  , √ σ 3 N1  |x|3

x ∈ R, N ≥ 1.

B.1

Taking Yi  wi − N δ−1 , where wi is defined by 2.1, yields MN − N δ SN   , N δ 1 − N δ−1    σ 2 N δ−1 1 − N δ−1 ,

B.2

 3  E Yi  ≤ 2N δ−1 . Thus, B.1 becomes     3cN −δ/2 MN − N δ   , ≤ x − Φx ≤ Prob    1  |x|3 N δ 1 − N δ−1 

B.3

for all N ≥ N0 for some N0 sufficiently large, depending on δ. For each N, k, and consider the event

AN,k, 

⎧ ⎪ ⎨ ⎪ ⎩



Mk;N − N

δ

N δ 1 − N δ−1 

⎫ ⎪  ⎬ . ∈ − N ,N ⎪ ⎭

B.4

Then, as N → ∞, we have Prob

1−δ N 

k1

AN,k,

−→ 1,

B.5

18

International Journal of Mathematics and Mathematical Sciences

provided that 1−δ N 

  Prob AcN,k, −→ 0,

B.6

k1

as N → ∞. Using B.3 gives      6cN −δ/2 Prob AcN,k, ≤  2 1 − Φ N 3 1  N     2 − N 2 −δ/2−3 N exp −  ≤ 6cN π 2

B.7

e.g., 32. Thus, the sum in B.6 is at most  6cN

1−3δ/2−3



2 1−δ− N exp π

 −

 N 2 , 2

B.8

and this is O1 provided that > 0 and > 1/3 − δ/2. Acknowledgments The authors would like to thank Ted Hill, Christoph Leuenberger, Daniel Stone, and the referees for numerous helpful comments. S. J. Miller was partially supported by NSF Grant no. DMS-0600848. References 1 F. Benford, “The law of anomalous numbers,” Proceedings of the American Philosophical Society, vol. 78, no. 4, pp. 551–572, 1938. 2 T. Hill, “The first-digit phenomenon,” American Scientists, vol. 86, pp. 358–363, 1996. 3 R. A. Raimi, “The first digit problem,” The American Mathematical Monthly, vol. 83, no. 7, pp. 521–538, 1976. 4 W. Hurlimann, “Benford’s law from 1881 to 2006,” preprint, http://arxiv.org/abs/math/0607168. 5 J. L. Brown Jr. and R. L. Duncan, “Modulo one uniform distribution of the sequence of logarithms of certain recursive sequences,” The Fibonacci Quarterly, vol. 8, no. 5, pp. 482–486, 1970. 6 P. Diaconis, “The distribution of leading digits and uniform distribution mod1,” The Annals of Probability, vol. 5, no. 1, pp. 72–81, 1977. 7 T. P. Hill, “A statistical derivation of the significant-digit law,” Statistical Science, vol. 10, no. 4, pp. 354–363, 1995. 8 A. Berger, L. A. Bunimovich, and T. P. Hill, “One-dimensional dynamical systems and Benford’s law,” Transactions of the American Mathematical Society, vol. 357, no. 1, pp. 197–219, 2005. 9 A. Berger and T. P. Hill, “Newton’s method obeys Benford’s law,” American Mathematical Monthly, vol. 114, no. 7, pp. 588–601, 2007. 10 D. Jang, J. U. Kang, A. Kruckman, J. Kudo, and S. J. Miller, “Chains of distributions, hierarchical Bayesian models and Benford’s law,” preprint, http://arxiv.org/abs/0805.4226. 11 A. V. Kontorovich and S. J. Miller, “Benford’s law, values of L-functions and the 3x  1 problem,” Acta Arithmetica, vol. 120, no. 3, pp. 269–297, 2005. 12 J. C. Lagarias and K. Soundararajan, “Benford’s law for the 3x  1 function,” Journal of the London Mathematical Society, vol. 74, no. 2, pp. 289–303, 2006.

S. J. Miller and M. J. Nigrini

19

13 S. J. Miller and M. J. Nigrini, “The modulo 1 central limit theorem and Benford’s law for products,” International Journal of Algebra, vol. 2, no. 1–4, pp. 119–130, 2008. 14 J.-C. Pain, “Benford’s law and complex atomic spectra,” Physical Review E, vol. 77, no. 1, Article ID 012102, 3 pages, 2008. 15 E. Costas, V. Lopez-Rodas, F. J. Toro, and A. Flores-Moya, “The number of cells in colonies of the ´ cyanobacterium Microcystis aeruginosa satisfies Benford’s law,” Aquatic Botany, vol. 89, no. 3, pp. 341– 343, 2008. 16 M. Nigrini and S. J. Miller, “Benford’s Law applied to hydrology data—results and relevance to other geophysical data,” Mathematical Geology, vol. 39, no. 5, pp. 469–490, 2007. 17 D. E. Knuth, The Art of Computer Programming, Volume 2: Seminumerical Algorithms, Addison-Wesley, Reading, Mass, USA, 3rd edition, 1997. 18 M. Nigrini, “Digital analysis and the reduction of auditor litigation risk,” in Proceedings of the Deloitte & Touche / University of Kansas Symposium on Auditing Problems, M. Ettredge, Ed., pp. 69–81, University of Kansas, Lawrence, Kan, USA, 1996. 19 M. Nigrini, “The use of Benford’s law as an aid in analytical procedures,” Auditing: A Journal of Practice & Theory, vol. 16, no. 2, pp. 52–67, 1997. 20 W. R. Mebane Jr., “Election forensics: the second-digit Benford’s law test and recent American presidential elections,” in Presented at the Election Fraud Conference, Salt Lake City, Utah, USA, September 2006. 21 M. Nigrini and S. J. Miller, “Data diagnostics using second order tests of Benford’s law,” preprint. 22 L. M. Leemis, B. W. Schmeiser, and D. L. Evans, “Survival distributions satisfying Benford’s law,” The American Statistician, vol. 54, no. 4, pp. 236–241, 2000. 23 H.-A. Engel and C. Leuenberger, “Benford’s law for exponential random variables,” Statistics & Probability Letters, vol. 63, no. 4, pp. 361–365, 2003. 24 R. S. Pinkham, “On the distribution of first significant digits,” Annals of Mathematical Statistics, vol. 32, pp. 1223–1230, 1961. 25 S. J. Miller and R. Takloo-Bighash, An Invitation to Modern Number Theory, Princeton University Press, Princeton, NJ, USA, 2006. 26 H. A. David and H. N. Nagaraja, Order Statistics, Wiley Series in Probability and Statistics, John Wiley & Sons, Hoboken, NJ, USA, 3rd edition, 2003. 27 R.-D. Reiss, Approximate Distributions of Order Statistics. With Applications to Nonparametric Statistics, Springer Series in Statistics, Springer, New York, NY, USA, 1989. 28 S. Lang, Undergraduate Analysis, Undergraduate Texts in Mathematics, Springer, New York, NY, USA, 2nd edition, 1997. 29 E. M. Stein and R. Shakarchi, Fourier Analysis: An Introduction, vol. 1 of Princeton Lectures in Analysis, Princeton University Press, Princeton, NJ, USA, 2003. 30 I. Gradshteyn and I. Ryzhik, Tables of Integrals, Series, and Products, Academic Press, New York, NY, USA, 5th edition, 1965. 31 V. V. Petrov, Limit Theorems of Probability Theory: Sequences of Independent Random Variables, vol. 4 of Oxford Studies in Probability, The Clarendon Press, Oxford University Press, New York, NY, USA, 1995. 32 W. Feller, An Introduction to Probability Theory and Its Applications. Vol. I, John Wiley & Sons, New York, NY, USA, 2nd edition, 1962.

Advances in

Operations Research Hindawi Publishing Corporation http://www.hindawi.com

Volume 2014

Advances in

Decision Sciences Hindawi Publishing Corporation http://www.hindawi.com

Volume 2014

Journal of

Applied Mathematics

Algebra

Hindawi Publishing Corporation http://www.hindawi.com

Hindawi Publishing Corporation http://www.hindawi.com

Volume 2014

Journal of

Probability and Statistics Volume 2014

The Scientific World Journal Hindawi Publishing Corporation http://www.hindawi.com

Hindawi Publishing Corporation http://www.hindawi.com

Volume 2014

International Journal of

Differential Equations Hindawi Publishing Corporation http://www.hindawi.com

Volume 2014

Volume 2014

Submit your manuscripts at http://www.hindawi.com International Journal of

Advances in

Combinatorics Hindawi Publishing Corporation http://www.hindawi.com

Mathematical Physics Hindawi Publishing Corporation http://www.hindawi.com

Volume 2014

Journal of

Complex Analysis Hindawi Publishing Corporation http://www.hindawi.com

Volume 2014

International Journal of Mathematics and Mathematical Sciences

Mathematical Problems in Engineering

Journal of

Mathematics Hindawi Publishing Corporation http://www.hindawi.com

Volume 2014

Hindawi Publishing Corporation http://www.hindawi.com

Volume 2014

Volume 2014

Hindawi Publishing Corporation http://www.hindawi.com

Volume 2014

Discrete Mathematics

Journal of

Volume 2014

Hindawi Publishing Corporation http://www.hindawi.com

Discrete Dynamics in Nature and Society

Journal of

Function Spaces Hindawi Publishing Corporation http://www.hindawi.com

Abstract and Applied Analysis

Volume 2014

Hindawi Publishing Corporation http://www.hindawi.com

Volume 2014

Hindawi Publishing Corporation http://www.hindawi.com

Volume 2014

International Journal of

Journal of

Stochastic Analysis

Optimization

Hindawi Publishing Corporation http://www.hindawi.com

Hindawi Publishing Corporation http://www.hindawi.com

Volume 2014

Volume 2014