ISSN:
1614-0176
Keywords:
Keywords: ECM algorithm, generalized logistic distribution, location-scale model, maximum likelihood estimation.
Source:
Springer Online Journal Archives 1860-2000
Topics:
Mathematics
Notes:
Summary: In financial and also in environmental time series one often encounters data with skew distributions. In these cases applying standard procedures derived under a normality assumption can lead to grossly misleading results. A versatile alternative consists in models based on a generalized logistic distribution, where skewness can be modeled and estimated with a skewness parameter. Maximum likelihood estimation under this model setting is considered and the existence of a sequence of consistent and asymptotically normal solutions of the likelihood equations is shown. The ECM algorithm is proposed for evaluation of estimates. The behaviour of the proposed procedure in estimating time series models and its appropriate use in model selection is investigated via simulations of a simple AR(1) model.
Type of Medium:
Electronic Resource
URL:
http://dx.doi.org/10.1007/s101820050005
Permalink