# Statistical Analysis of Human Parasitic Infections

Nikolaou, Vasileios V. N (2001) Statistical Analysis of Human Parasitic Infections. MSc(R) thesis, University of Glasgow.

Full text available as:

## Abstract

In longitudinal studies, measurements are taken over time or space on the same individual. A wide variety of well-established procedures exist for modelling these data such as t-tests, the use of summary measures, analysis of variance and the method of maximum likelihood. The latter seems to be the most appropriate approach since its estimates are consistent and efficient for large samples and it deals directly with problems of missing data. Here, much emphasis is given to the model proposed by A. Azzalini (1994), which is not only based on the method of maximum likelihood but it also incorporates an appropriate correlation structure for the measurements on a single individual across time. This model is applied in a real data example. It concerns a cohort study of about 1,100 individuals from China and Nigeria, where an infectious disease is widely spread. Treatment is given to control the prevalence and intensity of the disease. Our aim was to assess the way in which factors such as age and sex are related with the above outcomes, estimate the size of the effect of the treatment, as well as the time needed for prevalence and intensity to reach the pre-treatment level as a result of reinfection and make comparisons between two different types of treatment. However, a major problem arises due to the high frequency of missing data. In this case, a generalized linear model is fitted, which although it does not take into account the correlation over time, its methods of inference are based on well-founded theory under the assumption of independent errors. In addition, the use of smoothing techniques to explore trends in the data in a non parametric manner is described. This provides a means of modelling the data without making parametric assumptions. Finally, the methodology of generalized additive models is used to explore non-linear effects in a model comparison setting. Such models provide a means of checking more formally on linearity assumptions and also provide a way of modelling the data even when these effects are non-linear but can be assumed to be smooth. However, they have to be treated with care since their methods of inference are approximate.

Item Type: Thesis (MSc(R)) Masters Adviser: David Crompton Biostatistics, Epidemiology 2001 Enlighten Team glathesis:2001-76458 Copyright of this thesis is held by the author. 19 Nov 2019 14:19 19 Nov 2019 14:19 https://theses.gla.ac.uk/id/eprint/76458