Abstract
In a family of random variables, Taylor’s law or Taylor’s power law of fluctuation scaling is a variance function that gives the variance \(\sigma ^{2}>0\) of a random variable (rv) X with expectation \(\mu >0\) as a power of \(\mu \): \(\sigma ^{2}=A\mu ^{b}\) for finite real \(A>0,\ b\) that are the same for all rvs in the family. Equivalently, TL holds when \(\log \sigma ^{2}=a+b\log \mu ,\ a=\log A\), for all rvs in some set. Here we analyze the possible values of the TL exponent b in five families of infinitely divisible two-parameter distributions and show how the values of b depend on the parameters of these distributions. The five families are Tweedie–Bar-Lev–Enis, negative binomial, compound Poisson-geometric, compound geometric-Poisson (or Pólya-Aeppli), and gamma distributions. These families arise frequently in empirical data and population models, and they are limit laws of Markov processes that we exhibit in each case.
Similar content being viewed by others
Data Availability
There are no data associated with this paper.
References
Applebaum, D.: Lévy Processes and Stochastic Calculus. Cambridge Studies in Advanced Mathematics 116, 2nd edn. Cambridge University Press, Cambridge (2009)
Bar-Lev, S.K.: Independent, tough identical results: the class of Tweedie on power variance functions and the class of Bar-Lev and Enis on reproducible natural exponential families. Int. J. Stat. Probab. 9(1), 30–35 (2020). https://doi.org/10.5539/ijsp.v9n1p30
Bar-Lev, S., Enis, P.: Reproducibility and natural exponential families with power variance functions. Ann. Stat. 14(4), 1507–1522 (1986)
Bar-Lev, S., Stramer, O.: Characterizations of natural exponential families with power variance functions by zero regression properties. Probab. Theory Relat. Fields 76, 509–522 (1987)
Bar-Lev, S.K., Bshouty, D., Letac, G.: Natural exponential families and self-decomposability. Stat. Probab. Lett. 13, 147–152 (1992)
Barndorff-Nielsen, O.E., Stelzer, R.: The multivariate supOU stochastic volatility model. Math. Financ. 23(2), 275–296 (2013)
Bartlett, M.S.: Some notes on insecticide tests in the laboratory and in the field. J. R. Stat. Soc. 3(2), 185–194 (1936)
Beall, G.: Methods of estimating the population of insects in a field. Biometrika 30(3/4), 422–439 (1939)
Beall, G.: The transformation of data from entomological field experiments so that the analysis of variance becomes applicable. Biometrika 32, 243–262 (1942)
Bertoin, J.: Subordinators, Lévy processes with no negative jumps, and branching processes. http://www.maphysto.dk/publications/MPS-LN/2000/8.pdf (2000)
Bliss, C.I.: Statistical problems in estimating populations of Japanese beetle larvae. J. Econ. Entomol. 34(2), 221–232 (1941)
Brown, M., Cohen, J.E., de la Peña, V.: Taylor’s law, via ratios, for some distributions with infinite mean. J. Appl. Probab. 54(3), 1–13 (2017). https://doi.org/10.1017/jpr.2017.25
Brown, M., Cohen, J.E., Tang, C.F., Yam, S.C.: Taylor’s law of fluctuation scaling for semivariances and higher moments of heavy-tailed data. Proc. Natl. Acad. Sci. USA 118(46), e2108031118 (2021). https://doi.org/10.1073/pnas.2108031118
Cohen, J.E.: Stochastic population dynamics in a Markovian environment implies Taylor’s power law of fluctuation scaling. Theor. Popul. Biol. 93, 30–37 (2014). https://doi.org/10.1016/j.tpb.2014.01.001
Cohen, J.E.: Every variance function, including Taylor’s power law of fluctuation scaling, can be produced by any location-scale family of distributions with positive mean and variance. Theor. Ecol. 13, 1–5 (2020)
Cohen, J.E., Poulin, R., Lagrue, C.: Linking parasite populations in hosts to parasite populations in space through Taylor’s law and the negative binomial distribution. Proc. Natl. Acad. Sci. USA 114(1), E47–E56 (2016)
Cohen, J.E., Davis, R.A., Samorodnitsky, G.: Heavy-tailed distributions, correlations, kurtosis, and Taylor’s law of fluctuation scaling. Proc. R. Soc. A 476, 20200610 (2020). https://doi.org/10.1098/rspa.2020.0610
Comtet, L.: Analyse combinatoire. Tome 1. Presses Universitaires de France, Paris (1970)
Eisler, Z., Bartos, I., Kertesz, J.: Fluctuation scaling in complex systems: Taylor’s law and beyond. Adv. Phys. 57(1), 89–142 (2008). https://doi.org/10.1080/00018730801893043
Fracker, S.B., Brischle, H.A.: Measuring the local distribution of Ribes. Ecology 25(3), 283–303 (1944)
Harris, T.E.: The Theory of Branching Processes. Die Grundlehren der Mathematischen Wissenschaften. Springer, Prentice-Hall, Inc., Berlin (1963)
Hayman, B.I., Lowe, A.D.: The transformation of counts of the cabbage aphid (Brevicoryne brassicae (L.)). N. Z J. Sci. 4, 271–278 (1961)
Jørgensen, B.: Exponential dispersion models. J. R. Stat. Soc. B 49(2), 127–162 (1987)
Jørgensen, B.: The Theory of Dispersion Models. Chapman & Hall, London (1997)
Jørgensen, B., Kokonendji, C.C.: Discrete dispersion models and their Tweedie asymptotics. ASTA Adv. Stat. Anal. 100, 4378 (2016)
Jurek, Z.J.: Remarks on the self decomposability and new examples. Demonstr. Math. 34(2), 29–38 (2001)
Kendal, W.S.: A probabilistic model for the variance to mean power law in ecology. Ecol. Model. 80(23), 293–297 (1995)
Sato, K.-I.: Lévy Processes and Infinitely Divisible Distributions. Cambridge Studies in Advanced Mathematics, vol. 68. Cambridge University Press, Cambridge (2013)
Schoutens, W.: Lévy Processes in Finance: Pricing Financial Derivatives (Wiley Series in Probability and Statistics). Wiley, New York (2003)
Steutel, F.W., van Harn, K.: Infinite Divisibility of Probability Distributions on the Real Line. Chapman & Hall/CRC Pure and Applied Mathematics, London (2003)
Taylor, L.R.: Aggregation, variance and the mean. Nature 189(4766), 732–735 (1961)
Taylor, R.A.J.: Taylor’s Power Law: Order and Pattern in Nature. Elsevier Academic Press, Cambridge (2019)
van Harn, K., Steutel, F.W., Vervaat, W.: Self-decomposable discrete distributions and branching processes. Zeitschrift für Wahrscheinlichkeitstheorie und Verwandte Gebiete 61, 97–118 (1982)
Acknowledgements
We thank two reviewers for excellent constructive comments. T.H. acknowledges partial support from the “Chaire Modélisation mathématique et biodiversité” of Veolia-Ecole Polytechnique-MNHN-Fondation X, and support from the labex MME-DII Center of Excellence (Modèles mathématiques et économiques de la dynamique, de l’incertitude et des interactions, ANR-11-LABX-0023-01 project). This work was also funded by CY Initiative of Excellence (Grant “ Investissements d’Avenir” ANR- 16-IDEX-0008), Project “EcoDep” PSI-AAP2020-0000000013. The ECODEP project is organized by Paul Doukhan. We are grateful for this opportunity to collaborate.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The authors have no conflicts of interest associated with this paper.
Additional information
Communicated by Li-Cheng Tsai.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendix
Appendix
Here we briefly sketch that if X is SD, it is a weak limit of a pure-death branching process with immigration (if \(\mathbb {N}_{0}\)-valued), or a weak limit of a continuous-time Lévy-driven Ornstein-Uhlenbeck process (if continuous). The analysis concerns the rvs X obeying TL with \(a=0\). From the scaling transform (14) introducing \(a=\log \sigma _{1}^{2}\) , the modifications for \(a\ne 0\) could be readily obtained. In both discrete and continuous cases, a population that is randomly annihilated is randomly regenerated by the recurrent arrivals of random quantities of immigrants, yielding a stationary or invariant distribution of population size.
1.1 Discrete Self-decomposable rvs and Pure-Death Branching Processes with Immigration in Continuous Time
van Harn et al. [33] construct a regenerative process in continuous time that produces discrete SD distributions in the long run. Consider a continuous-time homogeneous compound Poisson process \(P_{r}(t)\), \(t\ge 0\), \( P_{r}(0)=0\), having rate \(r>0\), with pgf
where h(z), with \(h(0)=0\), is the pgf of the sizes of the clones or immigrant clusters arriving at the jump times of \(P_{r}(t)\). Let
be the pgf of a pure-death branching process started with one particle at \( t=0\). (More general subcritical branching processes could be considered.) This expression of \(\varphi _{t}(z)\) is easily seen to solve \(\overset{.}{ \varphi }_{t}(z)=f(\varphi _{t}(z))=1-\varphi _{t}(z)\), \(\varphi _{0}(z)=z\), as is usual for a pure-death continuous-time Bellman-Harris branching process [21] with affine branching mechanism \(f(z)=r_{d}(1-z)\) with fixed death rate \(r_{d}=1\). The distribution function of the lifetime of the initial particle is \(1-e^{-t}\). Let \(X_{t}\) with \(X_{0}=0\) be a random process counting the current size of some population for which a random number of individuals (determined by h(z)) immigrate at the jump times of \(P_{r}(t)\). Each newly arrived individual is independently and immediately subject to the pure death process above. We have
with \(\phi _{t}(0)={P}(X_{t}=0)=\exp \{-r\int _{0}^{t}(1-h(1-e^{-s}))ds\},\) the probability that the population is extinct at t. As \(t\rightarrow \infty \),
So \(X:=X_{\infty },\) as the limiting population size of this pure-death process with immigration, is a SD rv [33]. Define the rv \(X_{c}\) implicitly by requiring that \(X\overset{d}{=}cX^{\prime }+X_{c}\), where \( X^{\prime }\) is an iid copy of X and \(0<c<1\). Then
is a pgf. In such models typically, a decaying subcritical branching population is regenerated by a random number of incoming immigrants at random Poissonian times.
1.2 Continuous Self-decomposable rvs and Ornstein–Uhlenbeck Process in Continuous-Time
When X is continuous and SD, X is the limiting distribution of population size as \(t\rightarrow \infty \) of some Ornstein-Uhlenbeck process \(X_{t}\):
driven by the Lévy process \(\mathcal {L}_{t}\) for which
where \(\Phi _{0}( \lambda ) \) is the PLSt of an infinitely divisible rv appearing in the representation (7) of \(\Phi ( \lambda ) =Ee^{-\lambda X}\). See [26].
We now show that for a TweBLE rv with \(\alpha \in ( -\infty ,0) \), there is no \(L_{0}( \lambda ) =-\log \Phi _{0}( \lambda ) \) such that \(L_{0}^{\prime }( \lambda ) \) is completely monotone on \(( 0,\infty ) \). This result means that the TweBLE rv for \(\alpha \in ( -\infty ,0) \) is not SD, just infinitely divisible. Indeed, with \(L(\lambda )=-\log \Phi (\lambda )\),
with \(L_{0}^{\prime }(\lambda )>0\) only if \(\lambda>\lambda _{c}=-\theta /\alpha >0,\) so not in the full range \(\lambda \in (0,\infty )\). So \(\Phi _{0}(\lambda )\) is not completely monotone on \((0,\infty )\), and is therefore not an infinitely divisible PLSt. This Poisson-gamma regime for which the limiting distribution is a Poisson sum P of iid gamma-distributed clusters of size \(\Delta \) was studied by [19, p. 17, Sect. 3.3.2], who underline what they call its “impact inhomogeneity”: \(E(\Delta )=C(\alpha )\cdot E(P)^{-1/\alpha }\) for some constant \(C(\alpha )>0\).
By contrast, if \(\alpha \in (0,1)\), then \(L_{0}(\lambda )\) may be written as
where
is a tempered Lévy measure integrating \(1\wedge x\). The driving process \(\mathcal {L}_{t}\) of \(X_{t}\), with PLSt \(Ee^{-\lambda \mathcal {L} _{t}}=e^{-tL_{0}(\lambda )}\), is a subordinator and \(X=X_{\infty }\) is a SD TweBLE rv obtained as the limiting distribution of the corresponding Ornstein-Uhlenbeck process. One could extend this construction to other SD subordinated Lévy families, such as those in [1, 6, 10, 28, 29].
Rights and permissions
About this article
Cite this article
Cohen, J.E., Huillet, T.E. Taylor’s Law for Some Infinitely Divisible Probability Distributions from Population Models. J Stat Phys 188, 33 (2022). https://doi.org/10.1007/s10955-022-02962-y
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s10955-022-02962-y