Summary: An aspect of discrete data analysis: tting a
beta-binomial distribution to the hospitals data
P.M.E.Altham, Statistical Laboratory, University of Cambridge
January 30, 2002
Address Centre for the Mathematical Sciences, Wilberforce Road, Cam-
bridge CB3 OWB, UK, Fax no 01223-337956.
I was an invited speaker at the meeting
`The Design and Analysis of Potency Assays for Biotechnology Products'
at the National Institute for Biological Standards and Control, UK, October
Keywords Over-dispersion, beta-binomial, S-Plus, glm.
Statistical analysis for discrete data, particularly for probability models such as
the binomial, Poisson and multinomial, is by now very well understood, with a
wealth of suitable software. It can happen that the standard glm (generalized
linear modelling) software is not completely appropriate, since over-dispersion
is present, relative to the standard distributions such as the Poisson or the
binomial. Failure to take account of this over-dispersion, for example in tting
a model such as log(p=(1 p)) = + x (where the covariate x is the dose)