Steepness in Natural Exponential Families

The present paper studies and develops the notion of steepness in multivariate natural exponential families. Let F = \{P(m,F); m \in MF\} be a multidimensional natural exponential family parameterized by its domain of the means MF and let \overline{m} be an element of \partial MF the means domain boundary. A necessary and sufficient condition for the variance function VF is established so that the family F be steep at \overline{m} \in \partial MF. Some characteristic properties of a steep family are given. Also, we investigate the asymptotic behaviour of a steep family F at \overline{m}.

Steepness in Natural Exponential Families

The present paper studies and develops the notion of steepness in multivariate natural exponential families. Let F = \{P(m,F); m \in MF\} be a multidimensional natural exponential family parameterized by its domain of the means MF and let \overline{m} be an element of \partial MF the means domain boundary. A necessary and sufficient condition for the variance function VF is established so that the family F be steep at \overline{m} \in \partial MF. Some characteristic properties of a steep family are given. Also, we investigate the asymptotic behaviour of a steep family F at \overline{m}.

___

  • Barndorff-Nielsen O.: Information and Exponential Families in Statistical Theory. Wiley, New York (1978).
  • Brown, L.D.: Fundamentals of Statistical Exponential Families. Inst. of Math. Statist. Lecture Notes -Monograph Series (1986).
  • Casalis M.: The 2d+4 simple quadratic natural exponential families onRd. The Annals of Statistics. 24, 4, 1828-1854 (1994).
  • Del Castillo, J.: The truncated normal distribution: a non steep exponential family. Ann. Inst. Statist. Math. 46, 57-66 (1994).
  • Hassairi A.: La classiŞcation des familles exponentielles naturelles surRpar l’actiondu groupe linaire deRn+1.Comptes R endus de l’Acad´emie des Sciences. Paris, t.315, s´erie I, 210 (1992).
  • Hassairi A. and Lajmi S.: Riesz Exponential families on Symmetric cones. Journal of Theoretical Probability. 14, 4, 927-948 (2001).
  • Jİrgensen B. Martinez R. and Tsao M.: Asymptotic behaviour of the variance function. The Scandinavian Journal of Statistics. 21, 223-243 (1994).
  • Letac G.: A characterizationof the Wishart exponential families by aninvariance property. Journal of Theoretical Probability. 2, 1, 71-86 (1989).
  • Letac G.: Lectures on Natural exponential families and their variance functions. Monograf. Mat. 50. Instituto de Matem´atica pura e Applicada, Rio de Janeiro (1990).
  • Letac G. and Mora M.: Natural real exponential families with cubic variance function. The Annals of Statistics. 18, 1-37 (1990). Masmoudi A.:
  • Structures asymptotiques d’une fonction variance. Compte Rendus de l’Acad´emie des Sciences, Paris, t. 239, S´erie I, 177-182 (1999).
  • Morris C.N.: Natural exponential families with quadratic variance functions. The Annals of Statistics. 10, 1, 65-80 (1982).
  • Rockafellar T.: Convex Analysis. New Jersey. Princeton University Press (1970).
  • Seshadri V.: The inverse gaussian distribution. Clarendon Press. Oxford (1993).
  • Tweedie M.C.K.: Functions of a statistical variate with given means, with special reference to Laplaciandistribution. Proceedings of the Cambridge Philosophical Society , 43, 41-49 (1947).
  • Wedderburn, R.W.M.: Quasi likelihood functions, generalized linear models and the Gauss- Newtonmethod. Biometrika 61, 435-447 (1974). AŞf MASMOUDI
  • Laboratoire de Probabilit´e et Statistiques. Universit´e de Sfax. Facult´e des Sciences, B.P 802, Sfax-TUNISIA e-mail: AŞf.Masmoudi@fss.rnu.tn