Abstract
Given \(0<q<1,\) every absolutely continuous distribution can be described in two different ways: in terms of a probability density function and also in terms of a q-density. Correspondingly, it has a sequence of moments and a sequence of q-moments, if they exist. In this article, new conditions on the q-moment determinacy of probability distributions are derived. In addition, results related to the comparison of the properties of probability distributions with respect to the moment- and q-moment determinacy are presented.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
Due to the popularity of the q-calculus, numerous q-analogues of classical probability distributions have emerged, both for discrete and absolutely continuous cases. For example, there are q-binomial, q-Poisson, q-exponential, q-Erlang, and other q-distributions. These distributions play a significant role not only in the q-calculus itself, but also in various applications, primarily in theoretical physics. See, for example, [1, 6, 10, 14]. Comprehensive information concerning q-distributions is presented in [6] and, in this article, we follow the terminology and exposition of this monograph. Throughout the paper, \(q\in (0,1)\) is taken to be fixed. The q-integral defined by Jackson for \(0<a<b\) as
will be used along with the improper q-integral on \([0,+\infty )\) defined as
See [11, Sec. 19].
Definition 1.1
[6] Let P be a probability distribution with a distribution function F satisfying \(F(0)=0.\) A function f(t), \(t>0,\) is a q-density of P if
Correspondingly, the nth-order q-moment of P is
Clearly,
It has to be mentioned here that if P has a q-density f, then f is the q-derivative of the distribution function F, that is,
It is known ([11, Theorem 20.1]) that if \(F(0)= 0\), and it is continuous at 0, and then F can be represented in the form (1.1) and, therefore, possesses a q-density. In this paper, only probability distributions satisfying these conditions and possessing finite q-moments of all orders will be considered. The set of such distributions will be denoted by \({{\mathcal {A}}}.\) The q-moment problem in terms of q-densities has been studied in [17].
Notice that q-moments \(m_q(n;f)\) may also be obtained as the moments of a discrete distribution concentrated on \(\{q^j\}_{j\in {{\mathbb {Z}}}},\) whose probability mass function is given as
The moment problem for such discrete distributions was investigated in [2] by C. Berg, who explicitly found infinite families of distributions all possessing the same moments of all orders. These families can also be viewed as discrete Stieltjes classes, although the name “Stieltjes class” was suggested by Stoyanov ([18]) a few years after [2] had been published. The classical moment problem has deep connections with q-orthogonal polynomials, basic hypergeometric series, and theory of special functions. See, for example, [7, 8, 13] and the bibliography therein. The moment problems associated with various sequences of q-orthogonal polynomials have been investigated extensively by a large number of authors since an indeterminate moment problem implies non-uniqueness of the weight functions defining the respective inner product, like, for example, for the q-Laguerre, the discrete q-Hermite II, and Stieltjes-Wigert polynomials ([12, 3.21 and 3.27] and [2, Sect. 3 and 4]).
In the context of the q-calculus, the q-moments and strict q-moments were defined by G. Carnovale and T. Koornwinder in the form slightly different from (1.2). See [5, formula (1.2)] and [4, formula (2.3)]. In these works, the notion of the q-convolution on the line was studied, and it was found that the condition of q-moment determinacy is crucial for the commutativity of the q-convolution. Pertinent to this problem, certain conditions concerning the rate of growth of the q-moments were considered.
In this article, we stick with the notions and terminology of [6], where q-moments are defined for q-densities of continuous probability distributions by (1.2). Since the q-moments depend only on the values of a q-density on the sequence \(\{q^j\}_{j\in {{\mathbb {Z}}}},\) it is reasonable, therefore, to consider the following equivalence relation for functions on \((0, \infty ),\)
Definition 1.2
[17] A distribution \(P\in {{\mathcal {A}}}\) with a q-density f is q-moment determinate if \(m_q(n;f)=m_q(n;f_1)\) for all \(n\in {\mathbb {N}}_0\) implies that \(f \sim f_1.\) Otherwise, P is q-moment indeterminate.
It should be pointed out that every absolutely continuous distribution possessing finite moments of all orders can be examined from two different perspectives: those of moment determinacy and q-moment determinacy.
In [17], some conditions have been provided both for q-moment determinacy and indeterminacy in terms of the values \(f(q^{-j}).\) More precisely, it has been proved that
-
(i)
if
$$\begin{aligned} f(q^{-j})=o(q^{j(j+1)/2}), \quad j\rightarrow \infty , \end{aligned}$$(1.4)then P is q-moment determinate;
-
(ii)
if
$$\begin{aligned} f(q^{-j}) \geqslant Cq^{j(j+1)/2}, \quad j\geqslant 0, \end{aligned}$$(1.5)then P is q-moment indeterminate.
Statement (i) implies immediately that if a q-density f has a bounded support, then the distribution P is q-moment determinate.
In the sequel, the following q-analogue of the exponential function
is used as well as the q-exponential distribution with parameter \(\lambda >0,\) whose q-density is
For ample information on \(e_q(t)\) and the related distributions, we refer to [6, Section 1] and [11, Section 9].
In this work, new results on q-moment (in)determinacy are presented, both in terms of q-moments and q-density itself. Alternatively, it can be stated that some ‘checkable’ conditions for q-moment (in)determinacy are given. For the classical moment problem, an extensive review of such conditions can be found in [15]. The exact relation between the two moment problems is yet to be described. To this end, the present work is an attempt to connect these two aspects.
2 Statement of Results
Throughout the paper, the letter C with or without an index denotes a positive constant whose exact value does not need to be specified. We start with the assertion providing a condition for q-moment determinacy in terms of q-moments, in distinction to (1.4), where the values of the q-density are used.
Theorem 2.1
Let \(P\in \mathcal {A}\) have a q-density f and a sequence of q-moments \(\{m_q(n;f)\}_{n=0}^\infty \). If
then P is q-moment determinate.
To establish conditions for q-moment indeterminacy, we have to impose some restrictions on the behavior of a q-density, as in the next statement. In the theorem below, it is assumed that the sequence \(\{f(q^{-j})\}_{j\in {\mathbb {N}}_0}\) is log-concave.
Theorem 2.2
Let \(P\in \mathcal {A}\) have a q-density f and a sequence of q-moments \(\{m_q(n;f)\}_{n=0}^\infty \). If
and
then P is q-moment indeterminate.
Remark 2.1
It has to be pointed out that in both Theorems 2.1 and 2.2, the inequalities for A have to be strict. If \(A=\ln (1/q)/2,\) even if the q-density is subject to (2.3), the q-moment determinacy cannot be established without additional information. This is illustrated in Example 3.1 with the help of the q-gamma distribution whose q-density is equivalent to (3.3).
The next result provides a condition for the q-moment indeterminacy in situations not covered by the outcomes of Theorems 2.1 and 2.2. More precisely, an analogue of condition (1.5) for a subsequence \(\{q^{-mj}\}_{j=0}^\infty \) of \(\{q^{-j}\}_{j=0}^\infty \) is established.
Theorem 2.3
Let f be a q-density of a probability distribution \(P\in {{\mathcal {A}}}\). If, for a fixed positive integer m,
then the distribution P is q-moment indeterminate.
At this stage, the results related to an interrelation between the notions of moment- and q-moment determinacy will be presented. The next statement shows that the properties of a probability distribution with respect to moment- and q-moment determinacy may be diverse. Notice that the conditions of the next theorem are given in terms of the classical moments, while the conclusion is concerned with the q-moment problem.
Theorem 2.4
Let \(P\in {{\mathcal {A}}}\) and \(\{\mu _n\}_{n=1}^{\infty }\) be a sequence of its moments such that
If \(q<e^{-2a},\) then P is q-moment determinate. In particular, if \(q\in (0,1)\) and
then P is q-moment determinate.
The preceding results lead to the following
Theorem 2.5
There exist probability distributions which are simultaneously moment indeterminate and q-moment determinate. Moreover, there exist probability distributions which are moment indeterminate and q-moment determinate for each \(q\in (0,1).\)
3 Proofs of the Results
In this section, the notation \(M(r;f):=\max _{|z|=r} |f(z)|\), where f(z) is a function analytic in \(\{z:|z|=r\}\), will be used.
Proof of Theorem 2.1
Assume that there exists a q-density \(g\not \sim f\) such that \(m_q(n;f)=m_q(n;g)\) for all \(n\in {{\mathbb {N}}}_0,\) that is,
The existence of the q-moments implies that the Laurent series \(\sum _{j\in {{\mathbb {Z}}}} f(q^{-j})z^j\) and \(\sum _{j\in {{\mathbb {Z}}}} g(q^{-j})z^j\) converge in \({{\mathbb {C}}}^*={{\mathbb {C}}}{\setminus } \{0\}\) to \(\phi _1(z)\) and \(\phi _2(z),\) respectively, both of which are analytic in \({{\mathbb {C}}}^*.\) Now, we apply the following lemma proved in [17, Lemma 2.6]:
Lemma 3.1
Let \(\phi (z)=\sum _{j\in {{\mathbb {Z}}}} c_j z^j\) satisfy \(\phi (q^{-n})=0\) for all \(n\in {{\mathbb {N}}}.\) Then, for some \(C=C(q)>0, \) one has
By Lemma 3.1, for \(\phi =\phi _1-\phi _2,\) (3.1) holds. On the other hand,
whence
due to the assumption (2.1). Meanwhile, (3.1) yields
which contradicts (3.2). \(\square \)
Proof of Theorem 2.2
To prove the statement, it suffices to show that, under the conditions (2.2) and (2.3), the density f satisfies the condition (1.5). Consider
Here, \(\psi _1\) is a function analytic at \(\infty \) with \(\psi _1(\infty )=0,\) whence \(M(r;\psi _1)\rightarrow 0\) as \(r\rightarrow \infty .\) Consequently,
For a q-density f, one has \(m_q(n;f)=(1-q)\psi (q^{-(n+1)})=(1-q)M(q^{-(n+1)};\psi )\) for all \(n\in {{\mathbb {N}}}.\) Therefore,
To complete the proof, we refer to the result below by V. Boicuk and A. Eremenko.
Theorem 3.2
[3, Theorem 3] Let \(f(z)=\sum _{j=0}^\infty c_jz^j\) be an entire function such that \(|c_{j-1}c_{j+1}|\leqslant |c_j|^2\) and
Then,
Applying this result to \(\psi _2\) with \(c_j=f(q^{-j})\) and \(\beta =A/\ln ^2(1/q)\) yields
implying that, for \(A>\ln (1/q)/2, \) one has
for some \(C>0\) and j large enough. Since the sequence \(\{f(q^{-j})\}_{j\geqslant 0}\) does not vanish, it follows that this condition is satisfied for all \(j\in {\mathbb {N}}_0\). Thus, by (1.5) the distribution P is q-moment indeterminate. \(\square \)
Example 3.1
Using the q-exponential function given in (1.6), consider the q-gamma function of the form
as defined, for example, in [9, formula (4.1) with \(A=(1-q)^{-1}\)]. By a q-gamma distribution with a shape parameter \(\alpha >0\) and scaling parameter \(\lambda >0,\) it is meant a distribution whose q-density is equivalent to
Clearly, if \(\alpha =1,\) one recovers (1.7), while if \(\alpha =n\in {{\mathbb {N}}},\) one has the q-density of the n-stage q-Erlang distribution. See [6, Sec. 24] and [14]. Notice that f(t) satisfies condition (2.3) whatever \(\lambda >0\) is since
As the q-moments of the q-gamma distribution are
one derives the recurrence relation
and, hence,
where \((a;q)_j\) stands for the q-shifted factorial defined by
It is worth pointing out that, when \(\lambda = 1,\) the q-moments (3.4) coincide with the moment sequence associated with the q-Laguerre polynomials ([16, formula (2.13)], [2, formula (4.1)] and [8, formula (2.0.9)]), and that this moment sequence corresponds to an indeterminate Stieltjes moment problem. Explicit examples of distributions with these moments are provided in [2, Proposition 4.1] and [8, Propositions 2.1 and 2.2].
As for the q-moment determinacy, it can be easily obtained from (3.4) that, for all \(\lambda , \alpha >0,\)
that is, Theorems 2.1 and 2.2 are not applicable. Meanwhile, estimating, for \(j\geqslant 0,\)
one derives
By virtue of formulae (1.4) and (1.5), one concludes that a q-gamma distribution is q-moment determinate if and only if \(\lambda > q^{1-\alpha }/(1-q).\)
In particular, the q-exponential distribution with q-density (1.7) is q-moment determinate if and only if \(\lambda > 1/(1-q).\)
Proof of Theorem 2.3
Recall the well-known Euler identity [11, Section 9]:
Consider the entire function \(\phi _m(z)=\prod _{j=1}^\infty (1-q^{mj}z)\), for which it is clear that \(\phi _m(q^{-m(n+1)})=0\) for all \(n\in {{\mathbb {N}}}_0.\) By virtue of the Euler identity
Define the sequence \(\{c_j\}_{j\in {\mathbb {Z}}}\) in such a way that
where \(\alpha >0\) is selected so that \(c_j\geqslant 0\) for all \(j\in {{\mathbb {Z}}}.\) Such a selection is possible due to condition (2.4). Furthermore, using (3.5) and the fact that f is a q-density, one obtains that
Setting \(G(0)=0\) and \(G(q^j)=(1-q)\displaystyle \sum _{l=j}^\infty c_lq^l,\;j\in {\mathbb {Z}},\) and taking G(x) to be non-decreasing on each interval \([q^{j+1},q^j],\) one obtains a distribution function G, whose q-density g satisfies \(g(q^j)=c_j\) for all \(j\in {\mathbb {Z}}.\) Obviously, \(g \not \sim f\). Meanwhile, applying (3.5), one obtains:
Thus, P is q-moment indeterminate.\(\square \)
Note that the result cannot be derived from Theorem 2.2, although
Proof of Theorem 2.4
Let F be the distribution function and f be the q-density of P, respectively. Then one has:
by virtue of (1.3). Hence,
since \(q<e^{-2a}\) according to the assumption of the theorem. By Theorem 2.1, P is q-moment determinate.\(\square \)
Example 3.2
Let f(t) be a density of a generalized gamma distribution with parameters \(\alpha ,\) \(\beta ,\) \(\gamma >0,\) that is,
It is known—see [19, Section 11.4]—that the moments of the distribution are
and that it is moment indeterminate if and only if \(\gamma \in (0, 1/2)\). Since
we conclude by Theorem 2.4 that the generalized gamma distribution is q-moment determinate for all \(q\in (0,1)\) regardless of its parameter values.
Proof of Theorem 2.5
The assertion of the theorem is an immediate consequence of either Examples 3.1 or 3.2. The details are below.
(a) Example 3.1 shows that the q-exponential distribution possessing q-density (1.7) is q-moment determinate whenever \(\lambda >1/(1-q)\). Meanwhile, P is moment indeterminate for all \(\lambda >0.\) This can be established with the help of the Krein condition [19, Section 11, p.101]. Indeed, the density \(\rho \) of P can be estimated as
The bounds for \(e_q(-\lambda t)\) are derived from the next inequality obtained in [20, formula (2.6)], which holds for some positive constants \(C_1=C_1(q),\) \(C_2=C_2(q)\) and t large enough:
Thus,
Consequently, the Krein integral \(\displaystyle \int _{t_0}^{\infty } \frac{-\ln \rho (t^2)}{1+t^2} \, dt \) converges and, hence, the distribution P is moment indeterminate for all \(\lambda > 0\) and \(q\in (0,1).\)
(b) Example 3.2 also proves this theorem since, when \(\gamma \in (0,1/2),\) a generalized gamma distribution is moment indeterminate while being q-moment determinate for any given \(q\in (0,1).\) \(\square \)
References
Biedenharn, L.C.: The quantum group \(\text{ SU }_q\)(2) and a \(q\)-analogue of the boson operators. J. Phys. A Math. Gen. 22, L873–L878 (1989)
Berg, C.: On some indeterminate moment problems for measures on a geometric progression. J. Comput. Appl. Math. 99, 67–75 (1998)
Boicuk, V.S., Eremenko, A.E.: The growth of entire functions that are representable by Dirichlet series (Russian). Izv. Vysš. Učebn. Zaved. Matematika 5(156), 93–95 (1975)
Carnovale, G.: On the \(q\)-convolution on the line. Constr. Approx. 18, 309–341 (2002)
Carnovale, G., Koornwinder, T.H.: A \(q\)-analogue of convolution on the line. Methods Appl. Anal. 7(4), 705–726 (2000)
Charalambides, Ch.A.: Discrete \(q\)-Distributions. Wiley, Hoboken (2016)
Christiansen, J.S.: The moment problem associated with the Stieltjes–Wigert polynomials. J. Math. Anal. Appl. 277(1), 218–245 (2003a)
Christiansen, J.S.: The moment problem associated with the \(q\)-Laguerre polynomials. Constr. Approx. 19(1), 1–22 (2003b)
De Sole, A., Kac, V.G.: On integral representations of \(q\)-gamma and \(q\)-beta functions. arXiv:math/0302032vI (2003)
Jing, S.: The \(q\)-deformed binomial distribution and its asymptotic behaviour. J. Phys. A Math. Gen. 27, 493–499 (1994)
Kac, V., Cheung, P.: Quantum Calculus. Springer, Berlin (2002)
Koekoek, R., Swarttouw, R.F.: The Askey-scheme of hypergeometric orthogonal polynomials and its q-analogue. arXiv:math/9602214 (1996)
Koekoek, R., Lesky, P.A., Swarttouw, R.F.: Hypergeometric Orthogonal Polynomials and Their \(q\)-Analogues, Springer Monographs in Mathematics. Springer, Berlin (2010)
Kyriakoussis, A., Vamvakari, M.: Heine process as a \(q\)-analog of the Poisson process—waiting and interarrival times. Commun. Stat.-Theory Methods 46(8), 4088–4102 (2017)
Lin, G.D.: Recent developments on the moment problem. J. Stat. Distrib. Appl. 4, 5 (2017). https://doi.org/10.1186/s40488-017-0059-2
Moak, D.S.: The \(q\)-analogue of the Laguerre polynomials. J. Math. Anal. Appl. 81, 20–47 (1981)
Ostrovska, S., Turan, M.: \(q\)-Stieltjes classes for some families of \(q\)-densities. Stat. Probab. Lett. 146, 118–123 (2019)
Stoyanov, J.: Stieltjes classes for moment-indeterminate probability distributions. J. Appl. Probab. 41A, 281–294 (2004)
Stoyanov, J.: Counterexamples in Probability, 3rd edn. Dover Publications, New York (2013)
Zeng, J., Zhang, C.: A \(q\)-analog of Newton’s series, Stirling functions and Eulerian functions. Results Math. 25, 370–391 (1994)
Acknowledgements
The authors would like to extend their sincere gratitude to Prof. Alexandre Eremenko from Purdue University, USA for his valuable comments during the work on this paper. Also, appreciations go to Mr. P. Danesh from the Atilim University Academic Writing and Advisory Centre for his help in the preparation of the manuscript. Last but not least, we express our immense gratitude to the referees, whose valuable comments helped to improve the manuscript essentially.
Author information
Authors and Affiliations
Corresponding author
Additional information
Communicated by Anton Abdulbasah Kamil.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
About this article
Cite this article
Ostrovska, S., Turan, M. On the q-moment Determinacy of Probability Distributions. Bull. Malays. Math. Sci. Soc. 43, 3885–3896 (2020). https://doi.org/10.1007/s40840-020-00894-y
Received:
Revised:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s40840-020-00894-y