Abstract
Let F be the cumulative distribution function (CDF) of the base-q expansion \(\sum _{n=1}^\infty X_n q^{-n}\), where \(q\ge 2\) is an integer and \(\{X_n\}_{n\ge 1}\) is a stationary stochastic process with state space \(\{0,\ldots ,q-1\}\). In a previous paper we characterized the absolutely continuous and the discrete components of F. In this paper we study special cases of models, including stationary Markov chains of any order and stationary renewal point processes, where we establish a law of pure types: F is then either a uniform or a singular CDF on [0, 1]. Moreover, we study mixtures of such models. In most cases expressions and plots of F are given.
Similar content being viewed by others
Data Availability
NA.
References
Billingsley P (1965) Ergodic Theory and Information. Wiley
Billingsley P (1995) Probability and Measure. Wiley Series in Probability and Statistics, Wiley
Borel É (1909) Les probabilités dénombrables et leurs applications arithmétiques. Rend Circ Matem Palermo 27:247–271
Cantor G (1884) De la puissance des ensembles parfaits de points. Acta Math 4:381–392
Cesàro E (1906) Fonctions continues sans dérivées. Arch Math Phys 38:57–63
Cornean H, Herbst I, Møller J, Støttrup BB, Studsgaard KS (2022) Characterization of random variables with stationary digits. J Appl Prob 59:to appear. Avaliable at arXiv:2001.08492
de Amo E, Carrillo MD, Fernández-Sánchez J (2012) Singular functions with applications to fractal dimensions and generalized Takagi functions. Acta Appl Math 119:129–148
Denjoy A (1932) Sur quelgues points de la théorie des fonctions. C R Acad Sci Paris 194:44–46
Denjoy A (1934) Sur une fonction de Minkowski. C R Acad Sci Paris 198:44–47
de Rham G (1956) Sur quelques courbes définies par des équations fonctionnelles. Rend Sere Math Torino 16:101–113
Dovgoshey O, Martio O, Ryazanov V, Vuorinen M (2006) The Cantor function. Expo Math 24:1–37
Dym H (1968) On a class of monotone functions generated by ergodic sequences. Amer Math Mon 75:594–601
Harris TE (1955) On chains of infinite order. Pacific J Math 5:707–724
Ising E (1925) Beitrag zur theorie des ferromagnetismus. Z Phys 31:253–258
Jessen B, Wintner A (1935) Distribution functions and the Riemann zeta function. Trans Am Math Soc 38:48–88
Kairies H (1997) Functional equations for peculiar functions. Aequ Math 53:207–241
Kyeonghee J (2016) A construction of strictly increasing continuous singular functions. J Korean Soc Math Educ Ser B 23:21–34
Lyons R, Steif JE (2003) Stationary determinantal processes: Phase multiplicity, Bernoullicity, entropy, and domination. Duke Math J 120:515–575
Minkowski H (1904) Zur Geometrie der Zahlen, Verhandlungen des III. Internationalen Mathematiker-Kongresses in Heidelberg, 1904, pp. 164–173. (Gesammelte Abhandlungen von Hermann Minkowski. Bd. II. B. G. Teubner, Leipzig, 1911, pp. 43–52). Reprinted by Chelsea, New York, 1967
Okamoto H, Wunsch M (2007) A geometric construction of continuous, strictly increasing singular functions. Proc Japan Acad Ser A Math Sci 83(7):114–118. https://doi.org/10.3792/pjaa.83.114
Paradís J, Viader P, Bibiloni L (2007) Riesz-Nágy singular functions revisited. J Math Anal Appl 329:592–602
Paradís J, Viader P, Bibiloni L (2011) A new singular function. Am Math Mon 118(4):344–354. https://doi.org/10.4169/amer.math.monthly.118.04.344, https://www.tandfonline.com/doi/abs/10.4169/amer.math.monthly.118.04.344
Peres Y, Schlag W, Solomyak B (2000) Sixty years of Bernoulli convolutions. In: Brandt C, Graf S, Zähle M (eds) Fractal Geometry and Stochastics II, Birkhäuser Basel, vol 46, pp 39–65. https://doi.org/10.1007/978-3-0348-8380-1_2
Rao RR (1962) Relations between weak and uniform convergence of measures with applications. The Annals of Mathematical Statistics 33(2):659–680. http://www.jstor.org/stable/2237541
Riesz F, Sz-Nagy B (1955) Functional Analysis. Dover Publications
Salem R (1943) Some singular monotonic functions which are strictly increasing. Trans Amer Math Soc 53:427–439
Sánchez JF, Viader P, Paradís J, Carrillo MD (2012) A singular function with a non-zero finite derivative. Nonlinear Anal Theory Methods Appl 75(13):5010–5014. https://doi.org/10.1016/j.na.2012.04.015, http://www.sciencedirect.com/science/article/pii/S0362546X12001447
Sánchez JF, Viader P, Paradís J, Carrillo MD (2014) A singular function with a non-zero finite derivative on a dense set. Nonlinear Anal Theory Methods Appl 95:703–713. https://doi.org/10.1016/j.na.2013.10.001
Sánchez JF, Viader P, Paradís J, Carrillo MD (2016) A singular function with a non-zero finite derivative on a dense set with Hausdorff dimension one. J Math Anal Appl 434(1):713–728. https://doi.org/10.1016/j.jmaa.2015.09.036
Soshnikov A (2000) Determinantal random point fields. Russian Math Surveys 55:923–975
Takács L (1978) An increasing continuous singular function. Amer Math Monthly 85:35–37
Varjú PP (2018) Recent progress on Bernoulli convolutions. In: Mehrmann V, Skutella M (eds) Proceedings of the 7th European Congress of Mathematics, American Mathematical Society Bookstore, p 847–867. https://doi.org/10.4171/176-1/38
Wen L (1998) An approach to construct the singular monotone functions by using Markov chains. Taiwan J Math 2:361–368
Acknowledgements
This work was supported by The Danish Council for Independent Research | Natural Sciences, grant DFF - 10.46540/2032-00005B.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Competing Interests
The authors have no relevant financial or non-financial interests to disclose.
Additional information
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendix
Appendix
In this appendix we verify the theorems, the propositions, and the corollary in Sections 2–5 which remain to be proven, and we establish some related results. It is convenient to introduce the notation
for \(x_1,\dots ,x_n\in \{0,\dots ,q-1\}\).
1.1 Proof of Proposition 1
Let \(x=(0.x_1x_2\ldots )_q\in [0,1]\), where the non-terminating expansion is chosen if \(x\in \mathbb Q_q\). Then \(y=(0.y_1y_2\dots )_q\) is strictly less than x if and only if \(y_n<x_n\) for the first index n where \(y_n\ne x_n\), and thus by the law of total probability,
Hence (4) is verified and (4) implies (5).
1.2 Proof of Proposition 2
Let \(x=(0.x_1x_2\ldots )_q\in [0,1]\setminus \mathbb Q_q\) (by stationarity, F is continuous at x if \(x\in \mathbb Q_q\), cf. Remark 2). Then, by monotonicity of probabilities,
and so Proposition 2 follows immediately.
1.3 Proof of Proposition 3
The first part of Proposition 3 follows immediately from the mixture representation of F as given by (I)–(III) in Section 1.1, where c is the probability of obtaining the case (I) (i.e., \(F=cF_1+(1-c)\tilde{F}\) where \(\tilde{F}\) is a singular CDF on [0, 1]). For the second part of Proposition 3 we only consider the case \(m\ge 1\) as the case \(m=0\) follows from similar arguments. Then the following lemma will be useful.
Lemma 1
Let \(A\subset \mathbb {R}\) and suppose that \(f:A\mapsto \mathbb {R}\) is differentiable at \(x\in A\) and \(\{a_n\}_{n\ge 1},\{b_n\}_{n\ge 1}\) \(\subset A\setminus \{x\}\) are sequences converging to x such that there exists \(c>0\) with \(\vert a_n-b_n\vert \ge c\max \{\vert a_n-x\vert ,\vert b_n-x\vert \}\) for all \(n\in \mathbb {N}\). Then
Proof
We have
where the first inequality follows from the triangle inequality and the second from the assumption on \(\vert a_n-b_n\vert\). As the right hand side above goes to 0 for \(n\rightarrow \infty\), the proof is complete.
Note that in Lemma 1 it does not matter from which side the sequences \(\{a_n\}_{n\ge 1}\) and \(\{b_n\}_{n\ge 1}\) approach x. Thus, letting \(n\in \mathbb {N}\) and x be as in Proposition 3, we define \(a_n=(0.x_1\dots x_n \xi _1\dots \xi _m)_q+q^{-n-m}\) and \(b_n=(0.x_1\dots x_n \xi _1\dots \xi _m)_q\). Observe that \(a_n,b_n\in [0,1]\setminus \{x\}\). Then
and hence by Lemma 1,
Thus, since F is continuous at base-q fractions and both \(a_n\in \mathbb Q_q\) and \(b_n\in \mathbb Q_q\) for sufficiently large n, we get
whereby the proof is completed.
1.4 Proof of Corollary 1
We only prove Corollary 1 for \(m>1\), since the cases \(m=0\) and \(m=1\) follow from similar arguments. By Proposition 3,
This implies \(p(x_1,\dots ,x_n,\xi _1,\dots ,\xi _m)>0\) for n sufficiently large. Since also
we have
Thereby (7) is verified.
1.5 Proof of Proposition 4
The proof is straightforward when considering each of the cases \(x=0\), \(x=1\), \(x\in \mathbb Q_q\), and \(x\in (0,1)\setminus \mathbb Q_q\). For instance, suppose \(x=(0.x_1 x_2\dots )_q\in (0,1)\setminus \mathbb Q_q\). For \(n\in \mathbb {N}\), define \(y_n:=(0.x_1\dots x_n)_q\) and \(z_n:= y_n+q^{-n}\). Then \(y_n<x< z_n\) at least for sufficiently large n, and since \(y_n\in \mathbb Q_q\) and \(z_n\in \mathbb Q_q\), stationarity implies \(\mathrm P(X=y_n)=\mathrm P(X=z_n)=0\), and so \(F(z_n)-F(y_n)=p(x_1,\dots ,x_n)\). Thereby Proposition 4 is verified in the case where \(x=(0.x_1 x_2\dots )_q\) is a non-base-q fraction in (0, 1).
1.6 Proof of Theorem 1
Assume \(\{X_n\}_{n\ge 1}\) is a stationary Markov chain of order \(m-1\) for which F is not the uniform CDF on [0, 1]. Then it follows from (8) that there must exist \(\xi _1,\dots ,\xi _{m}\in \{0,\dots ,q-1\}\) such that \(\pi _{\xi _1,\dots ,\xi _{m}}\ne q^{-1}\). Let \(x=(0.x_1x_2\dots )_q\in (0,1)\setminus \mathbb Q_q\) where \(F'(x)\) exists. If \(F'(x)>0\), then we obtain a contradiction:
where the first equality follows from Corollary 1 and the second from the Markov property. Consequently, \(F'(x)=0\) and thus F is singular.
Using a notation as in Section 3, assume that \(\pi _{x_1,\ldots ,x_m}<1\) for all \(x_1,\dots ,x_m\in \{0,\dots , q-1\}\). Consider any \(x=(0.x_1x_2\dots )_q\in [0,1]\) and define
Then, for any integer \(n\ge m\), (9) gives
Hence, Proposition 2 gives that F is continuous at x. This completes the proof of Theorem 1.
1.7 Proof of Proposition 5
For any \(m\in \mathbb N\), denote the finite dimensional probabilities of \(\{X_n^{(m)}\}_{n\ge 1}\) by
for \(n\in \mathbb {N}\) and \(x_1,\ldots ,x_n\in \{0,\ldots ,q-1\}\). By construction of \(\{X_n^{(m)}\}_{n\ge 1}\), we have \(p_m(x_1,\dots ,x_n) =p(x_1,\dots ,x_n)\) whenever \(n\le m\). For any \(x=(0.x_1\ldots x_m)_q\), stationarity implies that \(\mathrm P(X=x)=0\), and so it follows from (4) that
Let \(x=(0.x_1x_2\dots )_q\in [0,1]\) be arbitrary. Combining (29) with the fact that \(F^{(m)}\) is non-decreasing gives
Here, by the continuity of \(F_3\), the left and the right hand side expressions of the inequalities converge to F(x) as \(m\rightarrow \infty\), so \(F^{(m)}\) converges pointwise to F (weak convergence). Hence, since F is a continuous CDF, we obtain (10), cf. Rao (1962).
1.8 Proof of Theorem 2
Let the situation be as in Theorem 2.
The case (I) follows immediately since then the \(X_n\)’s are independent and uniformly distributed on \(\{0,1\}\).
Suppose that \(\mathrm P(Z_1=k)=1\) for some \(k\in \mathbb N\). Then \(Z_0\) is uniformly distributed on \(\{1,\ldots ,k\}\). If \(\ell \in \{1,\ldots ,k\}\) and \(x=(0.x_1x_2\dots )_2=2^{-\ell }/(1-2^{-k})\), then \(x= \sum _{m=0}^\infty 2^{-\ell -km}\) and we get from (17) that \(p(x_1,\dots ,x_n)=1/k\). Thereby the case (II) is verified.
To show the case (III) assume first that \(Z_1\) is not geometrically distributed with mean 2. Equivalently,
for some \(m\in \mathbb {N}\). We show by contradiction that \(F'(x)=0\) for all \(x\in (0,1)\setminus \mathbb Q_2\) where \(F'(x)\) exists. So suppose that \(F'(x)>0\) for some \(x=(0.x_1x_2\dots )_2\in (0,1)\setminus \mathbb Q_2\). Let \(\xi _1=1\) and \(\xi _j=0\) for \(j\in \{2,\dots ,m+1\}\). Then, for any \(n\in \mathbb {N}\),
using in the first identity (17) and that \(Z_1,Z_2,\dots\) are identically distributed. This contradicts Corollary 1 and thus \(F'(x)=0\). In conclusion, F is singular if \(Z_1\) is not geometrically distributed with mean 2.
Suppose next that \(\mathrm {P}(Z_1=k)<1\) for all \(k\in \mathbb {N}\). If \(x\in [0,1]\) has finitely many digits equal to 1, then \(\mathrm P(X=x)=0\). If x has infinitely many digits equal to 1, then (17) gives
since \(\sup _{n\in \mathbb N}\mathrm P(Z_1=n)<1\). Hence, by Proposition 2, F is continuous.
Consequently, F is singular continuous in case (III).
Finally, Theorem 2(IV) follows immediately from Proposition 4 using (17) and the definition of the distribution of \(Z_0\).
1.9 Proof of Proposition 7
Let the situation be as in Proposition 7 and let G be given by (21). By Proposition 6(V), there exists \(c\in [0,1]\) such that \(c=G'(x)=F'(x)-\mathrm {P}(F_\Pi =F_1)\) for Lebesgue almost all \(x\in [0,1]\). We will show that \(c=0\) by contradiction, whereby Proposition 7 is verified.
Suppose that \(c>0\). For any \(j\in \mathbb {N}\), define \(U_j:= \{\pi \in \Omega \mid F_\pi \ne F_1\}\setminus V_j\). Since \(V_j\) increases to \(\{\pi \in \Omega \mid F_\pi \ne F_1\}\), there exists \(j_0\in \mathbb {N}\) such that \(P(\Pi \in U_{j_0})<c\). For \(x\in \mathbb {R}\), define
Then \(G=J+K\). Later on in this proof we will show that K, if not identically zero, must at least be singular, which implies that for Lebesgue almost all \(x\in [0,1]\), \(J'(x)=G'(x)=c>0\), hence by the definition of J, we have \(\mathrm {P}(\Pi \in U_{j_0})>0\). The function \(J/\mathrm {P}(\Pi \in U_{j_0})\) is a CDF satisfying (1), so by Proposition 6(V), \(c/\mathrm {P}(\Pi \in U_{j_0})=J'/ \mathrm {P}(\Pi \in U_{j_0})\le 1\) almost everywhere on [0, 1]. This is in contradiction with \(\mathrm {P}(\Pi \in U_{j_0})<c\), so \(c=0\).
Now let us show that \(K'=0\) almost everywhere. Clearly, we may assume that \(\mathrm {P}(\Pi \in V_{j_0})>0\) so that K is not identically zero. Then \(K/\mathrm {P}(\Pi \in V_{j_0})\) is a CDF satisfying (1) and therefore it is differentiable almost everywhere. Let A denote the set of points \(x\in [0,1]\) for which both \(K'(x)\) exists and the sequence \(\{f_{x,n}(\cdot )1[\cdot \in V_{j_0}]\}_{n\ge 1}\) converges pointwise to 0 as \(n\rightarrow \infty\) and is dominated by a constant. Then A has Lebesgue measure 1, since, by assumption A is the intersection of two sets of Lebesgue measure 1. Hence we can combine Proposition 6(V) with (20) and Lebesgue’s dominated convergence theorem to obtain
for all \(x=(0.x_1x_2\dots )_q\in A\). Thus K is singular.
1.10 Proof of Theorem 3
Theorem 3(I)–(II) follow directly from Proposition 6(III)–(IV).
When verifying Theorem 3(III) we assume \(m\ge 2\) and define without loss of generality \(\Omega\) as follows (the case \(m=1\) is simpler and follow similar lines as below). Denote the standard simplex in \(\mathbb {R}^n\) by
Consider the initial distribution \(\{\pi (x_1,\dots ,x_{m-1})\mid x_1,\ldots ,x_{m-1}\in \{0,\ldots ,q-1\}\}\) as a vector in \(\Delta _{q^{m-1}}\) and the collection of transition probabilities \(\{\pi _{x_1,\dots ,x_m}\mid x_1,\ldots ,x_m\in \{0,\ldots ,q-1\}\}\) as a \(q^{m-1}\)-tuple of vectors in \(\Delta _{q}\). Then \(\Delta _{q^{m-1}}\times \Delta _{q}^{q^{m-1}}\) can be identified with the collection of distributions for Markov chains with state space \(\{0,\ldots ,q-1\}\) and of order \(m-1\). Let \(\Omega\) be the subset of those elements of \(\Delta _{q^{m-1}}\times \Delta _{q}^{q^{m-1}}\) which correspond to stationary Markov chains of order \(m-1\), and let the \(\sigma\)-algebra on \(\Omega\) be induced by the Borel \(\sigma\)-algebra on \(\mathbb {R}^{q^{m-1}+q^{m}}\).
We now verify the requirements of Proposition 7 whereby Theorem 3(III) follows. Define
and let
Then \(V_1\subseteq V_2\subseteq \dots\). Further, let \(\pi ^*\in \Omega\) correspond to having all initial probabilities equal to \(q^{-m}\) and all transition probabilities equal to \(q^{-1}\) (so \(F_{\pi ^*}\) is the uniform CDF on [0, 1]). Then \(\pi ^*\) is the unique maximizer of \(\phi\) and \(\phi (\pi ^*)=1\). This follows e.g. with the use of Lagrange multipliers. Furthermore, it follows from (8) that \(F_\pi =F_1\) if and only if \(\pi =\pi ^*\). Hence
which is one requirement of Proposition 7.
To verify the other requirement of Proposition 7 we need some notation. For any \(x=(0.x_1x_2\ldots )_q\in [0,1]\setminus \mathbb Q_q\) and any \(n\in \mathbb N\), define
where \(n_{t_1,\dots ,t_j}(x_1,\dots ,x_k)\) is the number of times the string \(t_1,\dots ,t_j\) appears in the string \(x_1,\dots ,x_k\). Then the other requirement of Proposition 7 states that for all \(j\in \mathbb N\), Lebesgue almost all \(x=(0.x_1x_2\ldots )_q\in [0,1]\setminus \mathbb Q_q\), and all \(\pi \in \Omega\) we have that
converges to 0 as \(n\rightarrow \infty\) and is less than some number c(x, j). To verify this we recall that \((0.x_1x_2\dots )_q\in [0,1]\) is a normal number (in base q) if for all \(j\in \mathbb N\) and all \(t_1,\dots ,t_j\in \{0,\ldots ,q-1\}\),
Since the set of normal numbers in [0, 1] has Lebesgue measure 1 Borel (1909), we can assume that x is normal. By (31) there exists some \(n_j\in \mathbb N\) such that for all \(t_1,\dots ,t_m\in \{0,\dots ,q-1\}\),
whenever \(n\ge n_j\). Consequently, for all \(n\ge n_j\) and \(\pi \in V_j\),
where the second last inequality uses (30). Thereby the other requirement of Proposition 7 follows.
1.11 Proof of Theorem 4
Theorem 4(I)–(II) follow directly from Proposition 6(III)–(IV). Below we verify the requirements of Proposition 7 whereby Theorem 4(III) follows.
Define without loss of generality
and let the \(\sigma\)-algebra on \(\Omega\) be induced by the Borel \(\sigma\)-algebra on \(\ell ^1(\mathbb {N})\) (the space of absolutely summable sequences) equipped with the usual \(\ell ^1\)-norm. For any \(m\in \mathbb {N}\) and \(\pi \in \Omega\), define
and
so \(\pi ^*\in \Omega\) is the geometric distribution with mean 2. We now verify that
It is easily seen that \(\pi ^{(m)}\in \Omega\) is the unique maximizer of \(\phi _m\), and since for all \(\pi \in \Omega\), \(\phi _m(\pi )\) non-increases towards \(\phi (\pi )\) as \(m\rightarrow \infty\), it follows that
Furthermore, \(\Omega\) is a convex set and we claim that \(\phi\) is a strictly log-concave function: Consider any \(\alpha \in (0,1)\) and distinct \(\pi ,\tilde{\pi }\in \Omega\), so there exist some \(\pi _\ell \not =\tilde{\pi }_\ell\). Since \(\ln\) is strictly concave, we have \(\ln (\alpha \pi _\ell +(1-\alpha )\tilde{\pi }_\ell )>\alpha \ln \pi _\ell +(1-\alpha )\ln \tilde{\pi }_\ell\) and \(\ln (\alpha \pi _k+(1-\alpha )\tilde{\pi }_k)\ge \alpha \ln \pi _k+(1-\alpha )\ln \tilde{\pi }_k\) for \(k\not =l\), so
Consequently, \(\phi \le 1\) is strictly log-concave with
and so (32) follows.
Now, for any \(j\in \mathbb {N}\), define
Then \(V_1\subseteq V_2\subseteq \dots\). For all \(m\in \mathbb {N}\), a direct calculation gives \(\phi _m(\pi ^*)=2^{(m+2)/2^{m+1}}>1\), so \(\pi ^*\notin \bigcup _{j=1}^\infty V_j\). On the other hand, suppose \(\pi \in \Omega \setminus \pi \notin \bigcup _{j=1}^\infty V_j\). Then by (33) we can find an increasing sequence \(\{m_k\}_{k\ge 1}\) such that for all \(k\in \mathbb {N}\), we have \(\phi _{m_k}(\pi )>2^{-1/m_k}\), and so taking the limit as \(k\rightarrow\) we obtain \(\phi (\pi )\ge 1\). Therefore, by (32), \(\pi =\pi ^*\), so
From Theorem 2(I) and the definition of \(\pi ^*\) we obtain that \(F_\pi =F_1\) if and only if \(\pi =\pi ^*\), and hence
which is one requirement of Proposition 7.
For any \(j\in \mathbb {N}\) and any normal number \(x=(0.x_1x_2\dots )_q\in [0,1]\) (cf. (31)), define
By (31) there exists some \(n_j\in \mathbb N\) such that for all \(k\in \{1,\dots ,j\}\),
whenever \(n\ge n_j\). Hence, for all \(n\ge n_j\) and all \(\pi \in V_j\),
where the first inequality uses that \(\prod _{k=1}^m \pi _{k}^{n_k(x_1,\dots ,x_n)/n}\downarrow \prod _{k=1}^\infty \pi _{k}^{n_k(x_1,\dots ,x_n)/n}\) as \(m\rightarrow \infty\), the second inequality uses (34), and the last inequality uses (33). Therefore, for all \(n\ge n_j\) and all \(\pi \in V_j\), we have \(\phi _{x,n}(\pi )<1\). Furthermore, by definition, for all \(\pi \in V_j\) and \(n\in \mathbb {N}\), we have \(\phi _{x,n}(\pi )^n\le 2^n\). So, for any \(\pi \in \Omega\), we get that \(\{\phi _{x,n}^n1[\pi \in V_j]\}_{n\ge 1}\) is dominated by the constant \(2^{n_j}\) and converges pointwise to 0 as \(n\rightarrow \infty\). Thereby the other requirement of Proposition 7 holds, and so the proof of Theorem 4 is completed.
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Cornean, H., Herbst, I.W., Møller, J. et al. Singular Distribution Functions for Random Variables with Stationary Digits. Methodol Comput Appl Probab 25, 31 (2023). https://doi.org/10.1007/s11009-023-09989-y
Received:
Revised:
Accepted:
Published:
DOI: https://doi.org/10.1007/s11009-023-09989-y
Keywords
- Digit expansions of random variables in different bases
- Law of pure types
- Markov chain
- Mixture distribution
- Renewal process