Analysis of Dependent Data

• Credit value: 15 credits at Level 7
• Tutor: Georgios Papageorgiou

This module introduces you to more advanced theory, models and techniques for the analysis of data with complex dependency structure, for example hierarchical, repeated measures, longitudinal and spatial data. Practical examples, mainly in medical and biometric applications, will be presented, and there will be opportunities to explore the use of software for fitting such models using SAS and S+/R.

Indicative module syllabus

• Clustered observations and dependency, both ordered and exchangeable
• The need for appropriate techniques to account for correlation between observations
• Generalised least squares, including models for the covariance structure of the data
• Maximum likelihood and REML estimation
• The General Linear Mixed Model, including fixed and random effects
• Subject-specific models
• Use of the variogram as a means of identifying a suitable covariance structure
• The robust approach and the empirical 'sandwich' estimator
• Models for discrete data, including a review of the generalised linear model and an introduction to quasi-likelihood
• Marginal models using generalised estimating equations, GEEs and alternating logistic regressions for binary data
• Numeric integration using Gaussian quadrature

Time permitting, it may also be possible to briefly review classical methods (repeated measures ANOVA and MANOVA), and give an introduction to missing data issues.

By the end of this module, you should be able to demonstrate:

• substantial knowledge and understanding of the theory, models and techniques used for the analysis of data with complex structure and dependency, including repeated measures, longitudinal and spatial data
• knowledge and understanding of a range of methods for exploring and visualising structured data
• the ability to apply the theory to the statistical modelling and analysis of practical problems involving structured, dependent data, and to interpret results and draw conclusions in context
• the ability to use advanced statistical software for the analysis of complex statistical data
• the ability to incorporate the results of a technical analysis into clearly written report form that may be understood by a non-specialist.