Count data regression based on rich new families of distributions


The event is taking part on the Friday, Jan 19th 2018 at 11.00
Theme/s: Statistics
Location of Event: Alan Turing Building Room 306
This event is a: Public Seminar

Abstract: We propose a very flexible way of specifying new discrete distributions for use as an alternative to the Poisson and negative binomial distributions in regression models for count data. These discrete distributions are derived from renewal processes, i.e. distributions of the number of events by some time. Unlike the Poisson model, our models have time-dependent hazard functions. Any survival distribution can be used to describe the inter-arrival times between events, which gives a rich class of count processes with great flexibility for modelling both underdispersed and overdispersed data.

This framework is implemented in the R package Countr. It provides a number of built-in renewal count distributions and a model fitting function accompanied by standard methods for model exploration, diagnosis and prediction.
Distributions
based on user specified inter-arrival times are supported, as well.

This is joint work with Tarak Kharrat, Ian McHale and Rose Baker.

Report an error on this page

External Speakers

Georgi N. Boshnakov, (School of Mathematics, University of Manchester