Variance Calculation through Large Deviation Techniques - Inria - Institut national de recherche en sciences et technologies du numérique Access content directly
Reports (Research Report) Year : 2002

Variance Calculation through Large Deviation Techniques


In this paper, we show how to use the expression of entropy in order to calculate variances. Three cases are analyzed: independent and identically distributed (iid) variables, Markov chains (in discrete time) and jump Markov processes (in continuous time). This framework is valid far beyond these case studies, e.g. in transportation and telecommunication networks and likely in all models where the entropy is explicit. The method allows to derive the variance from the entropy function, which is a classical quantity in large deviations. Moreover, the entropy has often a rather simple expression (e.g. for networks). Here we show a closed formula expressing the variance in terms of derivatives of the entropy; by-products are also obtained, such as martingales, used in the proof of the central limit theorem. These results might be a good starting point for further developments, e.g. calculation of exact asymptotics (instead of logarithmic ones in large deviation) or solving entropy minimization problems.
Fichier principal
Vignette du fichier
RR-4441.pdf (612.13 Ko) Télécharger le fichier

Dates and versions

inria-00072147 , version 1 (23-05-2006)


  • HAL Id : inria-00072147 , version 1


Arnaud de La Fortelle. Variance Calculation through Large Deviation Techniques. [Research Report] RR-4441, INRIA. 2002. ⟨inria-00072147⟩
100 View
102 Download


Gmail Facebook Twitter LinkedIn More