Keywords: semi-Markov model, hazard rate, incidence and prevalence
The National Trachoma and Eye Health Program (1980) reports on the prevalance of otitis media (an infection that produces pus within the middle ear) in both aboriginal and non-aboriginal communities in Australia. The Program surveyed all aboriginal communities in Australia and attempted to contact all aborigines. Simultaneously, contact was made with non-aborigines usually living in the same or adjacent locations. Because of the high prevalence of infection in the aboriginal community only severe cases were classified as infected, virtually all of them suffering bursting of the ear drum and consequent scarring. It was thought that scarring could be used to identify those people who previously had had sever infections, but were not currently infected. So it was possible to classify subjects as (a) either not currently infected and no scarring, (b) currently infected or (c) not currently infected but one or more drums scarred. The data give the number of aborigines examined in various age intervals and the proportions classified as (a), (b) or (c).
|Age||Age interval (years)|
|Examined||Number of subjected examined|
|None||Proportion not currently infected and with no scarring|
|Current||Proportion currently infected|
|Past||Propotion not currently infected but with one or both drums scarred|
Data file (tab-delimited text)
|Royal Australian College of Ophthalmologists (1980. National Trachoma and Eye Health Program. Promail Printing Group, Artarmon, NSW.|
|McGilchrist, C. A., and Hills, L. J. (1991). A semi-Markov model for ear infection. Australian Journal of Statistics 33, 5-16.|
McGilchrist and Hills (1991) modelled incidence and recovery with a semi-Markov model in which incidence of the infection is allowed to be non-stationary.