Abstract: These mixture distributions are ones which have continuous piecewise linear density functions. These can be used to approximate any univariate distribution, but our focus will be to fit them to experimental data. We would like to use maximum likelihood methods to fit such distributions, but it seems that the literature does not provide good methods for making such fits. The current speaker will provide a new algorithm for making such a fit, and because of its simplicity, it can be easily implemented by practitioners in many applications. This will be demonstrated with some examples. As it is simple to sample from such mixture distributions, such a fitting can be useful for bootstrapping.
Biography: John graduated with a PhD in Pure Mathematics from University of Adelaide in 1975 with specialty in partial differential equations and functional analysis. He then moved to Applied Mathematics and worked in Stochastics with various applications in Mathematical Finance, Actuarial Science and Theoretical Statistics. In later years he has become more focused on statistical estimations with various applications including algorithms in machine learning and filtering. He holds an adjunct associate professorship at University of South Australia.
|Time:||5:30 pm - 7:30 pm|
|Location:||N132 (Engineering North Building),
South Australia 5001