Estimation of the Parameters of Extreme Value Distributions from Truncated Data Via the EM Algorithm

Estimation des paramètres de distributions des valeurs extrêmes à partir de données tronquées via l'algorithme EM

Tewfik Kernane (1) , Zohrh A. Raizah (2)

1 USTHB - Université des Sciences et de la Technologie Houari Boumediene = University of Sciences and Technology Houari Boumediene [Alger]
2 College of Sciences, King Khalid University

EM algorithm is used to obtain the maximum likelihood estimates for the parameters of extreme value distributions when the data are truncated. The method is used for the parameters of the type I least extreme value distribution (or Gumbel minimum) from right truncated data. Using transforms between the different types of extreme value distributions, the algorithm can be used to estimate the parameters of the Type I greatest extreme value distribution (Gumbel maximum) from left truncated data, for the two parameters Weibull distribution from right truncated data and for the Fréchet distribution from left truncated data. The algorithm is illustrated with simulated examples.

Domaines: Calcul [stat.CO]; Probabilités [math.PR]

Tewfik Kernane, Zohrh A. Raizah. Estimation of the Parameters of Extreme Value Distributions from Truncated Data Via the EM Algorithm. 2014. ⟨hal-00503252v2⟩