Abstract
We propose a novel complex-analytic method for sums of i.i.d. random variables that are heavy-tailed and integer-valued. The method combines singularity analysis, Lindelöf integrals, and bivariate saddle points. As an application, we prove three theorems on precise large and moderate deviations which provide a local variant of a result by Nagaev (Transactions of the sixth Prague conference on information theory, statistical decision functions, random processes, Academia, Prague, 1973). The theorems generalize five theorems by Nagaev (Litov Mat Sb 8:553–579, 1968) on stretched exponential laws \(p(k) = c\exp ( -k^\alpha )\) and apply to logarithmic hazard functions \(c\exp ( - (\log k)^\beta )\), \(\beta >2\); they cover the big-jump domain as well as the small steps domain. The analytic proof is complemented by clear probabilistic heuristics. Critical sequences are determined with a non-convex variational problem.
Similar content being viewed by others
Notes
For \(\beta \in (1,2]\), \(x q'(x)= \beta (\log x)^{\beta -1} \rightarrow \infty \) but the stronger condition \(x q'(x)/\log x\rightarrow \infty \) from Assumption 2.1(ii) fails. We suspect that this restriction is technical and could be lifted with more detailed estimates, but a proof or disproof is beyond this article’s scope.
References
Arratia, R., Barbour, A.D., Tavaré, S.: Logarithmic Combinatorial Structures, EMS Monographs in Mathematics. European Mathematical Society (EMS), Zürich (2003)
Armendáriz, I., Grosskinsky, S., Loulakis, M.: Zero-range condensation at criticality. Stoch. Proc. Appl. 123, 3466–3496 (2013)
Denisov, D., Dieker, A.B., Shneer, V.: Large deviations for random walks under subexponentiality: the big-jump domain. Ann. Probab. 36, 1946–1991 (2008)
Doetsch, G.: Theorie und Anwendung der Laplace-Transformationen. Springer, Berlin (1937)
Embrechts, P., Klüppelberg, C., Mikosch, T.: Modelling Extremal Events. For Insurance and Finance, Applications of Mathematics (New York), vol. 33. Springer, Berlin (1997)
Ercolani, N.M., Jansen, S., Ueltschi, D.: Random partitions in statistical mechanics. Electron. J. Probab. 19(82), 1–37 (2014)
Feller, W.: An Introduction to Probability Theory and its Applications, vol. 2. Wiley, New York (1966)
Flajolet, P., Gerhold, S., Salvy, B.: Lindelöf representations and (non-)holonomic sequences. Electron. J. Comb. 17, R3 (2010)
Flajolet, P., Sedgewick, R.: Analytic Combinatorics. Cambridge University Press, Cambridge (2009)
Ibragimov, I.A., Linnik, Y.U.: Independent and Stationary Sequences of Random Variables. Wolters-Noordhoff Publishing, Groningen (1971)
Lindelöf, E.: Le calcul des résidus et ses applications à la théorie des fonctions, Collection de monographies sur la théorie des fonctions, publiée sous la direction de M. Émile Borel. Gauthier-Villars, Paris, 1905. Reprinted by Gabay, Paris (1989)
Nagaev, A.V.: Local limit theorems with regard to large deviations when Cramér’s condition is not satisfied. Litov. Mat. Sb. 8, 553–579 (1968). Selected Transl. in Math. Stat. Probab. 11, 249–278 (1973)
Nagaev, A.V.: Letter to the editor: a remark concerning the article “Local limit theorems with regard to large deviations”. Litov. Mat. Sb. 10, 206 (1970). Selected Transl. in Math. Stat. Probab. 11, 279 (1973)
Nagaev, S.V.: Large deviations for sums of random variables. In: Transactions of the Sixth Prague Conference on Information Theory, Statistical Decision Functions, Random Processes (Tech. Univ., Prague, 1971), pp. 657–674. Academia, Prague (1973)
Nagaev, S.V.: Large deviations of sums of independent random variables. Ann. Probab. 7, 745–789 (1979)
Pollard, H.: The representation of \(\text{ e }^{-x^\lambda }\) as a Laplace integral. Bull. Am. Math. Soc. 52, 908–910 (1946)
Rozovskiĭ, L.V.: Probabilities of large deviations on the whole axis. Teor. Veroyatnost. i Primenen. 38, 79–109 (1993). Transl. in Theory Probab. Appl. 38, 53–79 (1993)
Acknowledgements
The authors wish to thank the Laboratoire Jean Dieudonné of the University of Nice and the Institut Henri Poincaré (during the program of the spring 2013 organized by M. Esteban and M. Lewin) for their kind hospitality and for the opportunity to discuss this project. S. J. thanks V. Wachtel for pointing out Nagaev’s article [14].
Author information
Authors and Affiliations
Corresponding author
Additional information
Supported by NSF Grant DMS-1212167.
Appendices
Appendix A. Proofs of Lemmas 2.14 and 2.15
Proof of Lemma 2.14
The proof of Assumption 2.1 is straightforward and left to the reader. The function \(q(\zeta ) = \zeta ^\alpha \) is analytic in \(\mathrm {Re}\,\zeta >0\) and \(p(\zeta )= \exp (- \zeta ^\alpha )\) satisfies, for all \(k \in \mathbb {N}\),
Since \(\alpha \in (0,1)\), we have \(\cos \frac{\alpha \pi }{2} >0\) and Eq. (A.1) shows that \(|\zeta ^k p(\zeta )|\) is integrable along \(\mathrm {Re}\,\zeta =1/2\) and that \(p(\zeta )\) grows slower than any exponential \(\exp (\varepsilon |\zeta |)\). This proves Assumption 2.2.
The equation \(q''(x_n^*) = -1/(n\sigma ^2)\) can be solved explicitly. \(N_n^*\) and \(N_n^{**}\) are best determined with the scaling relation (2.24). They have already been determined in [12], we omit the proof. For the insensitivity scale, we notice that
which goes to zero if and only if \(N_n \gg n^{-1/(2-2\alpha )}\). \(\square \)
Proof of Lemma 2.15
The function \(q(x) = - \log c +(\log x)^\beta \) is clearly smooth on \((1,\infty )\). Then \(q'(x) = \beta (\log x)^{\beta -1}/ x\) and as \(x\rightarrow \infty \),
Assumption 2.1 is easily checked. For Assumption 2.2 we note \(q(\zeta ) = c + (\log \zeta )^\beta \) is analytic in \(\mathrm {Re}\,\zeta >1\). Fix \(b>1\) and write \(\zeta = r \exp ( \mathrm{i}\theta )\). As \(|\zeta | \rightarrow \infty \) along \(\mathrm {Re}\,\zeta = b\) i.e. \(\zeta = b + \mathrm{i}y\), the argument \(\theta \) goes to \(\pm \pi /2\) and we have
conditions(i) and (ii) in Assumption 2.2 are easily checked. Assumption 2.3(iii) follows from a computation similar to (A.3). Set \(y_0 = \sqrt{r^2 - b^2}\). We have for \(\zeta = b + \mathrm{i}y\), \(y\ge y_r\), uniformly in r,
hence
which proves Assumption2.3(i). Next let \(\zeta \in \mathbb {C}\) with \(\mathrm {Re}\,\zeta >1\), write \(\zeta = r \exp (\mathrm{i}\theta )\), then
and for large r and \(\theta \in (0,\pi /2)\),
and so Assumption 2.3(ii) holds.
We now turn to the asymptotic behavior of the sequences \(x_n^*\), \(N_n^*\) and \(N_n^{**}\). Since \(q''(x_n) = -\frac{1}{n\sigma ^2}\), it is clear that \(x_n \rightarrow \infty \) as \(n \rightarrow \infty \). The equation is
Consequently,
The last bracket is asymptotically equal to 1 and we get the expression for \(x_n^*\). Next, we have from (2.10)
The last asymptotics follows from (A.10).
We now turn to \(N_n^{**}\). It is asymptotically given by the solution of the equations
Equation (A.14) is equivalent to
The relevant solution is
It follows that \(N_n - x_n \sim \frac{\beta n \sigma ^2}{N_n} (\log x_n)^{\beta -1}\). We insert this in (A.13); using \(\log x_n \sim \log N_n\), we get
The relevant solution is
Only the term \(\log n\) matters in the last bracket and the result follows.
The last part of the lemma on insensitivity sequence is shown in [3, Section 8.3], the proof is therefore omitted. \(\square \)
Appendix B. Bivariate Hessian
As explained in Step 4 of the proof outline, the Hessian at \((t_n,\zeta _n)\) has determinant \(-1 + o(1)\) and is a saddle point of \(\Phi _n(t,\zeta )\), considered as a function of two real variables \(t,\zeta >0\). In order to get rid of off-diagonal elements in the Hessian and to give all eigenvalues the same sign, we take complex \(\zeta \) and reparametrize, as sketched in Step 5.
Lemma B.1
Let \(\zeta (t)\) be the unique solution of \((\partial _\zeta \Phi _n)(t,\zeta ) =t - q'(\zeta ) = 0\). Set
Then \((\nabla F_n)(t_n,0) =0\),
Note that \(\zeta _n = \zeta (t_n)\), so as \(n\rightarrow \infty \)
Proof
We have
At \(t=t_n\), \(s=0\), we have \(\zeta (t) = \zeta _n\) and \((\nabla F_n)(t_n,0) = \nabla \Phi _n(t_n,\zeta _n) =0\). For the Hessian, we compute
By definition of \(\zeta (t)\),
It follows that \(\partial _t \partial _s F_n(t,0) =0\), and
We solve for \(\zeta '(t)\) in Eq. (B.4), insert into Eq. (B.5), and obtain the formula for \(\beta (t)\). \(\square \)
Rights and permissions
About this article
Cite this article
Ercolani, N.M., Jansen, S. & Ueltschi, D. Singularity Analysis for Heavy-Tailed Random Variables. J Theor Probab 32, 1–46 (2019). https://doi.org/10.1007/s10959-018-0832-2
Received:
Revised:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10959-018-0832-2
Keywords
- Local limit laws
- Large deviations
- Heavy-tailed random variables
- Asymptotic analysis
- Lindelöf integral
- Singularity analysis
- Bivariate steepest descent