- Profª M. Salomé Cabral – CEAUL/ DEIO (FCUL)
- FCUL – Campo Grande – Bloco C6 Piso 4 – Sala: 6.4.30 – 16:00h
- Quinta-feira, 3 de Dezembro de 2015
In many cancer studies and in much clinical research, repeated observations of response variable are taken over time for each individual in one or more treatment groups. Such research is commonly referred as “longitudinal” study in which the main aim is to compare treatment groups. Data from a mammary tumour experiment are analysed where the outcome is the number of tumours taken over time and the goal is to compare three diets used to study the influence of lipids on the development of cancer. The analysis of longitudinal count data poses some diculties in particular the repeated measures of each vector of responses are likely to be correlated and the autocorrelation structure for the repeated data plays a signicant role in the estimation of regression parameters. Several approaches have been proposed and two of them have been considered to analyse these data. The generalized estimation equations (GEE) method and the approach based on maximum likelihood estimation where the serial dependence is assumed to be of Markovian type (MML). In both cases the conclusion is that the log of the number of tumours increases quadratically and that the diet of low fat is the most appropriate to delay the development of the number of tumours. For GEE approach the estimates were obtained through the function geeglm in the R package geepack. The function cold in the R package cold is used to obtain the MML estimates.