505
Views
1
CrossRef citations to date
0
Altmetric
Research Article

Poisson degenerate central moments related to degenerate Dowling and degenerate r-Dowling polynomials

, &
Pages 583-597 | Received 12 Jul 2022, Accepted 24 Aug 2022, Published online: 02 Sep 2022

ABSTRACT

Degenerate Dowling and degenerate r-Dowling polynomials were introduced earlier as degenerate versions and further generalizations of Dowling and r-Dowling polynomials. The aim of this paper is to show their connections with Poisson degenerate central moments for a Poisson random variable with a certain parameter and with Charlier polynomials.

MATHEMATICS SUBJECT CLASSIFICATIONS:

1. introduction and preliminaries

In recent years, studying various degenerate versions of many special polynomials and numbers, which began with the paper [Citation1] by Carlitz, received regained interests of some mathematicians and many interesting results were discovered (see [Citation2–8] and the references therein). They have been explored by employing several different tools such as combinatorial methods, generating functions, p-adic analysis, umbral calculus techniques, differential equations, probability theory and analytic number theory.

Degenerate Dowling and degenerate r-Dowling polynomials were introduced earlier as degenerate versions and further generalizations of Dowling and r-Dowling polynomials. The aim of this paper is to show their connections with Poisson degenerate central moments for a Poisson random variable with a certain parameter and with Charlier polynomials.

The outline of this paper is as follows. In Section 1, we recall the Stirling numbers of the first and second kinds, Bell polynomials, the degenerate exponential functions, the degenerate Stirling numbers of the first and second kinds, the degenerate Bell polynomials, the Poisson random variable with parameter α, and the Charlier polynomials. In addition, we remind the reader of the Whitney numbers of the first and second kinds, Dowling polynomials, the degenerate Whitney numbers of the second kind, the degenerate Dowling polynomials, the degenerate r-Whitney numbers of the second kind and the degenerate r-Dowling polynomials. Section 2 is the main result of this paper. In the following, assume that X is the Poisson random variable with mean αm. In Theorem 2.1, we show that the Poisson degenerate central moment E[(mX+1)n,λ] is equal to Dm,λ(n,α), where Dm,λ(n,x) is the degenerate Dowling polynomial. In Theorem 2.2, we express the same Poisson degenerate central moment in terms of the degenerate Bell polynomials. In Theorem 2.4, we deduce that E[(mX+r)n,λ] is equal to Dm,λ(r)(n,α), where Dm,λ(r)(n,x) is the degenerate r-Dowling polynomial. We express the same in terms of the degenerate Bell polynomials in Corollary 2.5, and of the degenerate r-Whitney numbers of the second kind and the Bell polynomials in Theorem 2.6. Furthermore, it is represented by the Charlier polynomials and the degenerate Stirling numbers of the second kind in Theorems 2.10 and 2.11. In the rest of this section, we recall the facts that are needed throughout this paper.

It is well known that the stirling numbers of the first kind are defined as (1) (x)n=k=0nS1(n,k)xk,(n0),(see [914]),(1) where (x)0=1, (x)n=x(x1)(xn+1), (n1).

The Stirling numbers of the second kind are given by (2) xn=k=0nS2(n,k)(x)k,(n0),(see [8,15]).(2)

From (Equation1) and (Equation2), we note that (3) 1k!(log(1+t))k=n=kS1(n,k)tnn!,(3) and (4) 1k!(et1)k=n=kS2(n,k)tnn!,(k0),(see [7,12,15,16]).(4)

The Bell polynomials are defined by (5) ex(et1)=n=0ϕn(x)tnn!,(see [12,15,16]).(5)

Thus, by (Equation5), we get (6) ϕn(x)=k=0nS2(n,k)xk,(n0),(see [1,6,15]).(6)

In [Citation2], the degenerate exponentials are defined by (7) eλx(t)=(1+λt)xλ=n=0(x)n,λtnn!,eλ(t)=eλ1(t),(0λR),(7) where (x)0,λ=1,(x)n,λ=x(xλ)(x2λ)(x(n1)λ),(n1),(see [7]). Let logλt be the compositional inverse of eλ(t) such that logλ(eλ(t))=eλ(logλt)=t.

Then we have (8) logλ(1+t)=n=1λn1(1)n,1λn!tn,(see [2]).(8)

In view of (Equation1) and (Equation2), the degenerate Stirling numbers of the first kind, S1,λ(n,k), and the degenerate Stirling numbers of the second kind, S2,λ(n,k), are defined by (9) (x)n=k=0nS1,λ(n,k)(x)k,λ,(n0),(9) and (10) (x)n,λ=k=0nS2,λ(n,k)(x)k,(n0)(see [2]).(10)

Note that limλ0logλ(1+t)=log(1+t),limλ0eλx(t)=ext, limλ0S1,λ(n,k)=S1(n,k), and limλ0S2,λ(n,k)=S2(n,k), where (n,k0).

As degenerate versions of (Equation3) and (Equation4) and from (Equation9) and (Equation10), we note that (11) 1k!(eλ(t)1)k=n=kS2,λ(n,k)tnn!,(k0),(11) and 1k!(logλ(1+t))k=n=kS1,λ(n,k)tnn!,(see [2]), where eλ(t) and logλ(1+t) are respectively given by (Equation7) and (Equation8).

The degenerate Bell polynomials are defined by (12) ex(eλ(t)1)=n=0ϕn,λ(x)tnn!,(see [2,5,6]).(12)

Thus, by (Equation11) and (Equation12) and as a degenerate version of (Equation6), we get (13) ϕn,λ(x)=k=0nS2,λ(n,k)xk,(n0),(see [57]).(13)

A random variable X is a real valued function defined on a sample space. If X takes any value in a countable set, then X is called a discrete random variable. For a discrete random variable X, the probability mass function p(a) of X is defined by p(a)=P{X=a},(see [1719]). A random variable X taking on one of the values 0,1,2, is said to be the Poisson random variable with parameter α(>0), which is denoted by XPoi(α), if the probability mass function of X is given by (14) p(i)=P{X=i}=eααii!, i=0,1,2,,(see [1719]).(14)

For n1, the quantity E[Xn] of the Poisson random variable X with parameter α(>0), which is called the nth moment of X, is given by E[Xn]=i=0inp(i)=eαi=0ini!αi=ϕn(α),(see [1719]). As is well known, the Charlier polynomials Cn(x;α) are defined by (15) eαt(1+t)x=n=0Cn(x;α)tnn!,(see [15]),(15) where x,t,αR.

Thus, by (Equation15), we get (16) Cn(x;α)=l=0n(k=ln(nk)(1)nkαnkS1(k,l))xl.(16)

A finite lattice L is geometric if it is a finite semimodular lattice which is also atomic. Dowling constructed an important finite geometric lattice Qn(G) out of a finite set of n elements and a finite group G of order m, called Dowling lattice of rank n over a finite group of order m. If L is the Dowling lattice Qn(G) of rank n over a finite group G of order m, then the Whitney numbers of the first kind VQn(G)(n,k) and the Whitney numbers of the second kind WQn(n,k) are respectively denoted by Vm(n,k) and Wm(n,k). The Whitney numbers Vm(n,k) and Wm(n,k) satisfy the following Stirling number-like relations: (17) (mx+1)n=k=0nWm(n,k)mk(x)k,(17) mn(x)n=k=0nVm(n,k)(mx+1)k,(n0),(see [4,20,21]). For n0, Dowling polynomials are given by (18) Dm(n,x)=k=0nWm(n,k)xk,(see [3,20,21]).(18)

Recently, Kim-Kim considered a degenerate version of (Equation17), namely the degenerate Whitney numbers of the second kind defined by (19) (mx+1)n,λ=k=0nWm,λ(n,k)mk(x)k,(k0),(see [3]).(19)

Thus, by (Equation19), we get (20) eλ(t)1k!(eλm(t)1m)k=n=kWm,λ(n,k)tnn!.(20)

In [Citation4], the degenerate Dowling polynomials are defined by (21) eλ(t)exm(eλm(t)1)=n=0Dm,λ(n,x)tnn!.(21)

As a degenerate version of (Equation18) and from (Equation20) and (Equation21), we get (22) Dm,λ(n,x)=k=0nWm,λ(n,k)xk,(see [3,4]).(22)

A further generalization of degenerate Whitney numbers of the second kind, Kim-Kim introduced the degenerate r-Whitney numbers of the second kind given by (23) (mx+r)n,λ=k=0nWm,λ(r)(n,k)mk(x)k,(n,r0),(see [3,4]).(23)

In view of (Equation22), they defined the degenerate r-Dowling polynomials given by (24) Dm,λ(r)(n,x)=k=0nWm,λ(r)(n,k)xk,(n0),(see [3,4]).(24)

From (Equation24), we can show that the generating function of the degenerate r-Dowling polynomials is given by (25) eλr(t)exm(eλm(t)1)=n=0Dm,λ(r)(n,x)tnn!,(see [3,4]).(25)

2. Poisson degenerate central moments related to degenerate Dowling and degenerate r-Dowling polynomials

Let X be the Poisson random variable with mean αm. Then we consider the Poisson degenerate central moments given by E[(mX+1)n,λ], (n0). We observe from (Equation14) that (26) E[eλmX+1(t)]=k=0eλmk+1(t)p(k)=eλ(t)eαmk=0eλmk(t)1k!(αm)k=eλ(t)eαmeαmeλm(t)=eλ(t)eαm(eλm(t)1)=n=0Dm,λ(n,α)tnn!.(26) On the other hand, by (Equation7), we get (27) E[eλmX+1(t)]=n=0E[(mX+1)n,λ]tnn!.(27) Therefore, by (Equation26) and (Equation27), we obtain the following theorem.

Theorem 2.1

Let XPoi(αm). Then we have E[(mX+1)n,λ]=Dm,λ(n,α),(n0).

Note that E[(mX+1)n]=limλ0E[(mX+1)n,λ]=limλ0Dm,λ(n,α)=Dm(n,α),(n0).

Let XPoi(αm). It is not difficult to show that (28) (x+y)n,λ=k=0n(nk)(x)k,λ(y)nk,λ,(n0).(28)

By (Equation28), we get (29) E[(mX+1)n,λ]=k=0n(nk)(1)nk,λE[(mX)k,λ].(29)

From (Equation14), we have (30) n=0E[(mX)n,λ]tnn!=E[eλmX(t)]=eαmn=0eλmn(t)(αm)nn!=eαm(eλm(t)1)=eαm(eλm(mt)1)=n=0ϕn,λm(αm)mntnn!.(30) Comparing the coefficients on both sides of (Equation30), we have (31) E[(mX)n,λ]=ϕn,λm(αm)mn,(n0).(31)

By (Equation13), we get (32) ϕn,λm(αm)=k=0nS2,λm(n,k)(αm)k.(32)

Therefore, by (Equation29), (Equation31) and (Equation32), we get (33) E[(mX+1)n,λ]=k=0n(nk)(1)nk,λmkϕk,λm(αm)=k=0n(nk)(1)nk,λmkj=0kS2,λm(k,j)(αm)j=j=0nαjk=jn(nk)(1)nk,λmkjS2,λm(k,j).(33) Therefore, by (Equation33), we obtain the following theorem.

Theorem 2.2

For n0, let XPoi(αm). Then we have E[(mX+1)n,λ]=k=0n(nk)(1)nk,λmkϕk,λm(αm)=j=0nαjk=jn(nk)(1)nk,λmkjS2,λm(k,j).

When m = 1, we have n=0D1,λ(n,α)tnn!=n=0E[(X+1)n,λ]tnn!=E[eλX+1(t)]=eλ(t)eαn=0eλn(t)n!αn=eλ(t)eα(eλ(t)1)=eα(eλ(t)1)+ddαeα(eλ(t)1)=n=0(ϕn,λ(α)+ddαϕn,λ(α))tnn!. Thus, we have (34) ϕn,λ(α)+ddαϕn,λ(α)=D1,λ(n,α)=E[(X+1)n,λ],(n0).(34)

On the other hand, by (Equation13), we have (35) ddαϕn,λ(α)=ddαk=0nS2,λ(n,k)αk=k=1nkS2,λ(n,k)αk1=k=0n1(k+1)S2,λ(n, k+1)αk.(35) Therefore, by (Equation34) and (Equation35), we obtain the following corollary.

Corollary 2.3

For n0, let XPoi(α). Then we have E[(X+1)n,λ]=D1,λ(n,α)=k=0n((k+1)S2,λ(n,k+1)+S2,λ(n,k))αk.

For r0, let X be the Poisson random variable with mean αm. Then we have (36) E[eλmX+r(t)]=k=0eλmk+r(t)p(k)=eλr(t)eαmk=0eλmk(t)k!(αm)k=eλr(t)eαmeαmeλm(t)=eλr(t)eαm(eλm(t)1)=n=0Dm,λ(r)(n,α)tnn!.(36) The left hand side of (Equation36) is given by (37) E[eλmX+r(t)]=n=0E[(mX+r)n,λ]tnn!.(37)

Therefore, by (Equation36) and (Equation37), we obtain the following theorem.

Theorem 2.4

For m, r0, let XPoi(αm). Then we have E[(mX+r)n,λ]=Dm,λ(r)(n,α),(n0).

Note that Dm(r)(n,α)=limλ0E[(mX+r)n,λ]=E[(mX+r)n]. By (Equation28), we get (38) E[(mX+r)n,λ]=k=0n(nk)(r)nk,λE[(mX)k,λ]=k=0n(nk)(r)nk,λmkϕk,λm(αm).(38) Therefore, by (Equation38), we obtain the following corollary.

Corollary 2.5

For n, r0, let XPoi(αm). Then we have Dm,λ(r)(n,α)=E[(mX+r)n,λ]=k=0n(nk)(r)nk,λmkϕk,λm(αm).

When r = 1, we have E[(mX+1)n,λ]=Dm,λ(1)(n,α)=Dm,λ(n,α).

From (Equation23), we have (39) Dm,λ(r)(n,α)=E[(mX+r)n,λ]=k=0nWm,λ(r)(n,k)mkE[(X)k].(39)

By (Equation3), we get (40) n=0E[(X)n]tnn!=E[(1+t)X]=k=0E[Xk]1k!(log(1+t))k=k=0E[Xk]n=kS1(n,k)tnn!=n=0(k=0nS1(n,k)E[Xk])tnn!.(40) Since XPoi(αm), from (Equation14), we have (41) k=0E[Xk]tkk!=E[eXt]=eαmk=0ekt(αm)kk!=eαm(et1)=k=0ϕk(αm)tkk!.(41) Thus, by (Equation41), we get (42) E[Xk]=ϕk(αm),(k0).(42)

From (Equation40) and (Equation42), we have (43) E[(X)n]=k=0nS1(n,k)E[Xk]=k=0nS1(n,k)ϕk(αm).(43)

By (Equation39) and (Equation43), we get (44) Dm,λ(r)(n,α)=E[(mX+r)n,λ]=k=0nWm,λ(r)(n,k)mkE[(X)k]=k=0nWm,λ(r)(n,k)mkj=0kS1(k,j)ϕj(αm)=j=0n(k=jnWm,λ(r)(n,k)mkS1(k,j))ϕj(αm).(44) Therefore, by (Equation44), we obtain the following theorem.

Theorem 2.6

For n0, let XPoi(αm). Then we have Dm,λ(r)(n,α)=E[(mX+r)n,λ]=j=0n(k=jnWm,λ(r)(n,k)mkS1(k,j))ϕj(αm).

By Theorem 2.2, we get (45) Dm,λ(r)(n,α)=E[(mX+r)n,λ]=k=0n(nk)(r1)nk,λE[(mX+1)k,λ]=k=0n(nk)(r1)nk,λj=0kαjl=jk(kl)(1)kl,λmljS2,λm(l,j)=j=0nαj{k=jnl=jk(nk)(kl)(r1)nk,λ(1)kl,λmljS2,λm(l,j)}.(45) On the other hand, by (Equation24), we get (46) Dm,λ(r)(n,α)=j=0nαjWm,λ(r)(n,j).(46)

Therefore, by (Equation45) and (Equation46), we obtain the following theorem.

Theorem 2.7

For n, j0, we have Wm,λ(r)(n,j)=k=jnl=jk(nk)(kl)(r1)nk,λ(1)kl,λmljS2,λm(l,j).

From (Equation46) and (Equation38), we note that (47) j=0nαjWm,λ(r)(n,j)=Dm,λ(r)(n,α)=E[(mX+r)n,λ]=l=0n(nl)(r)nl,λmlϕl,λm(αm)=l=0n(nl)(r)n1,λmlj=0l(αm)jS2,λm(l,j)=j=0nαj(l=jn(nl)(r)nl,λmljS2,λm(l,j)).(47) Therefore, by comparing the coefficients on both sides of (Equation47), we obtain the following theorem.

Theorem 2.8

For n,j0, we have Wm,λ(r)(n,j)=l=jn(nl)(r)nl,λmljS2,λm(l,j).

We recall from (Equation15) (see also (Equation16)) that Charlier polynomials Cn(x;α) are given by (48) eαt(1+t)x=n=0Cn(x;α)tnn!,(n0 and x,α, tR).(48)

Let us take x = 0. Then we have (49) eαt=n=0Cn(0;α)tnn!.(49)

Replacing t by 1eλm(t) in (Equation49), we get (50) eα(eλm(t)1)=k=0Ck(0;α)(1)k1k!(eλm(t)1)k.(50)

Thus, by (Equation12) and (Equation50), we get (51) n=0ϕn,λm(α)mntnn!=eα(eλm(mt)1)=eα(eλm(t)1)=k=0Ck(0;α)(1)k1k!(eλm(mt)1)k=k=0Ck(0;α)(1)kn=kS2,λm(n,k)mntnn!=n=0(mnk=0nCk(0;α)(1)kS2,λm(n,k))tnn!.(51) Therefore, by (Equation51), we obtain the following theorem.

Theorem 2.9

For n0, let XPoi(α), Then we have ϕn,λm(α)=E[(X)n,λm]=k=0nCk(0;α)(1)kS2,λm(n,k).

Let us take x = 1 in (Equation48). Then we have (52) eαt(1+t)=n=0Cn(1;α)tnn!.(52)

Replacing t by eλm(t)1 and α by αm in (Equation52), we get (53) n=0Dm,λ(m)(n,α)tnn!=eλm(t)eαm(eλm(t)1)=k=0Ck(1;αm)(eλm(t)1)kk!=k=0Ck(1;αm)1k!(eλm(mt)1)k=n=0(k=0nCk(1;αm)mnS2,λm(n,k))tnn!.(53) Comparing the coefficients on both sides of (Equation53), we have (54) Dm,λ(m)(n,α)=k=0nCk(1;αm)mnS2,λm(n,k),(n0).(54)

Note that (55) E[(mX+r)n,λ]=E[(mX+m+rm)n,λ]=k=0n(nk)(rm)nk,λE[(mX+m)k,λ]=k=0n(nk)(rm)nk,λj=0kCj(1;αm)S2,λm(k,j)mk=j=0nCj(1;αm)k=jn(nk)(rm)nk,λS2,λm(k,j)mk.(55) Therefore, by (Equation54) and (Equation55), we obtain the following theorem.

Theorem 2.10

For n0, let XPoi(αm). Then we have E[(mX+m)n,λ]=Dm,λ(m)(n,α)=k=0nCk(1;αm)mnS2,λm(n,k). Furthermore, we have Dm,λ(r)(n,α)=E[(mX+r)n,λ]=j=0nCj(1;αm)k=jn(nk)(rm)nk,λS2,λm(k,j)mk.

Replacing t by (eλm(t)1), α by αm, and x by rm in (Equation48), we have (56) eαm(eλm(t)1)eλr(t)=k=0Ck(rm;αm)1k!(eλm(t)1)k=k=0Ck(rm;αm)1k!(eλm(mt)1)k=k=0Ck(rm;αm)n=kS2,λm(n,k)mntn!=n=0(mnk=0nCk(rm;αm)S2,λm(n,k))tnn!.(56) On the other hand, by (Equation25), we get (57) eλr(t)eαm(eλm(t)1)=n=0Dm,λ(r)(n,α)tnn!.(57)

Therefore, by (Equation56) and (Equation57), we obtain the following theorem.

Theorem 2.11

For n0, let XPoi(αm). Then we have Dm,λ(r)(n,α)=E[(mX+r)n,λ]=mnk=0nCk(rm;αm)S2,λm(n,k).

3. Conclusion

In recent years, studying various degenerate versions of many special polynomials and numbers received regained interests of some mathematicians and many interesting results were discovered. Degenerate Dowling and degenerate r-Dowling polynomials were introduced earlier as degenerate versions and further generalizations of Dowling and r-Dowling polynomials.

Assume that X is the Poisson random variable with mean αm. We showed that the Poisson degenerate central moment E[(mX+1)n,λ] is equal to Dm,λ(n,α) and to an expression involving the degenerate Bell polynomials, respectively in Theorems 2.1 and 2.2. We deduced that E[(mX+r)n,λ] is equal to Dm,λ(r)(n,α) in Theorem 2.4. We expressed the same in terms of the degenerate Bell polynomials in Corollary 2.5 and of the degenerate r-Whitney numbers of the second kind and the Bell polynomials in Theorem 2.6. Furthermore, it is represented by the Charlier polynomials and the degenerate Stirling numbers of the second kind in Theorems 2.10 and 2.11.

As one of our future projects, we would like to continue to study degenerate versions of certain special polynomials and numbers and their applications to physics, science and engineering as well as mathematics.

Ethics approval and consent to participate

The authors declare that there is no ethical problem in the production of this paper.

Consent for publication

The authors want to publish this paper in this journal.

Acknowledgments

The author would like to thank the referees for the detailed and valuable comments that helped improve the original manuscript in its present form. Also, The authors thank Jangjeon Institute for Mathematical Sciences for the support of this research.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by the Basic Science Research Program, the National Research Foundation of Korea (NRF-2021R1F1A1050151).

References

  • Carlitz L. Degenerate Stirling, Bernoulli and Eulerian numbers. Utilitas Math. 1979;15:51–88.
  • Kim DS, Kim T. A note on a new type of degenerate Bernoulli numbers. Russ J Math Phys. 2020;27(2):227–235.
  • Kim T, Kim DS. Degenerate Whitney numbers of the first and second kinds of Dowling lattices. Russ J Math Phys. 2022;29(2):81–90.
  • Kim T, Kim DS. Degenerate r-Whitney numbers and degenerate r-Dowling polynomials via boson operators. Adv in Appl Math. 2022;140: 102394. (Paper No. 102394):21 pp.
  • Kim T, Kim DS. Some identities on truncated polynomials associated with degenerate Bell polynomial. Russ J Math Phys. 2021;28(3):342–355.
  • Kim T, Kim DS, Dolgy DV. On partially degenerate Bell numbers and polynomials. Proc Jangjeon Math Soc. 2017;20(3):337–345.
  • Kim T, Kim DS, Jang LC, et al. Representations of degenerate Hermite polynomials. Adv in Appl Math. 2022;139: 102359. (Paper No. 102359):18 pp.
  • Kim T, Kim DS, Kim HK. Normal ordering of degenerate integral powers of number operator and its applications. Appl Math Sci Eng. 2022;30(1):440–447.
  • Araci S. A new class of Bernoulli polynomials attached to polyexponential functions and related identities. Adv Stud Contemp Math. 2021;31(1):195–204.
  • Bayad A, Chikhi J. Apostol-Euler polynomials and asymptotics for negative binomial reciprocals. Adv Stud Contemp Math. 2014;24(1):33–37.
  • Brillhart J. Mathematical notes: note on the single variable bell polynomials. Amer Math Monthly. 1967;74(6):695–696.
  • Comtet L. Advanced combinatorics: the art of finite and infinite expansions. Dordrecht: Reidel; 1974.
  • Djordjevic GB, Milovanovic GV. Special classes of polynomials. Leskovac: University of Nis, Faculty of Technology. 2014. p. 211 pp.
  • Simsek Y. Identities and relations related to combinatorial numbers and polynomials. Proc Jangjeon Math Soc. 2017;20(1):127–135.
  • Roman S. The umbral calculus. New York: Academic Press. Inc. (Harcourt Brace Jovanovich Publishers); 1984. (Pure and Applied Mathematics; 111).
  • Carlitz L. Some remarks on the Bell numbers. Fibonacci Quart. 1980;18(1):66–73.
  • Girardin V, Limnios N. Applied probability: from random experiments to random sequences and statistics. Cham: Springer; 2022.
  • Grimmett GR, Stirzaker DR. Probability and random processes. Oxford: Oxford University Press; 2020.
  • Ross SM. Introduction to probability models. London: Academic Press; 2019.
  • Dowling TA. A class of geometric lattices bases on finite groups. J Combin Theory Ser B. 1973;14:61–86.
  • Dowling TA, Wilson RM. Whitney number inequalities for geometric lattices. Proc Amer Math Soc. 1975;47:504–512.