Brief course descriptions

Compulsory Courses
Analysis of continuous data
In this course we lay the foundations of statistical modelling through (linear) regression.Based on a sample of observed continuous outcomes accompanied by a general set of predictors,  we explain and predict the outcome from the covariates. We pay special attention to interpretation and the concepts of confounding and interaction. Under a set of assumptions we draw inference under the model and  perform an analysis of (co)variance. We provide tools to build the model, to check assumptions  and to extend the model when needed.  We further provide guidance on study design and sample size calculations, all supported by hands on PC practicals. A broad range of applications is well served by this methodology or its extensions.

Categorical Data Analysis
This course enables students to understand and apply the most frequently used methods of categorical data analysis. After briefly introducing some discrete distributions and the maximum likelihood framework, various statistical models for analyzing binary, ordinal and nominal responses are presented. Among the models covered are loglinear models, logistic regression, and poisson regression. Students learn about large sample methods as well as exact small sample mehods to test common hypotheses in the context of categorical data analysis.

Statistical Inference
In statistical inference, we lay out some fundamental ways in which statisticians can use obtained data to derive information about an underlying, 'true' distribution function. While the course deals with some of the mathematical foundations of statistics, the goal is to offer the student a practical set of tools which he/she can use to understand common approaches to data analysis, but also when tackling more exotic, out-of-the ordinary situations

Optional Courses
Bayesian Statistics
In this course, the students are introduced to the principles of Bayesian estimation. It is shown how Bayesian inference is different from the frequentist (classical) approach. The goal of the course is that students are able to apply Bayesian techniques in relevant applications and that they acquire the skills to interpret the obtained results in a meaningful way.

Data Mining and Big Data

The goal of the course is to familiarize the students with the most important methods to extract information from large databases in a statistical way. The students are expected to learn how to use these techniques correctly in applications and they acquire the skills to interpret obtained results in a statistically correct manner. The students will also be introduced to big data and the problems this might impose.

Analysis of Clustered and Longitudinal Data 
This course discusses regression analysis of continuous and discrete outcomes that are correlated within independent units or groups: for instance, test scores from children who are grouped into classes and schools; or repeatedly measured outcome data for independent individuals. Basic tools for the analysis of correlated data will be explained, such as random effect models and generalised estimating equations.

Causality and Missing Data
The first part (2/3) of this course discusses techniques for estimating the effect of a (possibly time-varying) exposure, treatment, policy or intervention on an outcome from observational studies in which treated and untreated individuals may fail to be comparable due to confounding. Questions will be addressed such as `how much of this gene's effect on lung cancer is mediated via smoking?’ and `what is the effect of hospital-acquired infection on mortality?’. Techniques for causal inference will be explained, such as propensity score analysis, mediation analysis, marginal structural models and instrumental variables analysis. The second part (1/3) of this course discusses multiple imputation for the analysis of incomplete data sets.

Experimental Design
The course content is closely related to the theory and practice of linear statistical models. Although the design phase of a study appears prior to the experimentation and statistical analysis phases, a design cannot be constructed without knowing how the data, that will arise from the designed study, will be analyzed. A good design is necessary to make the statistical analysis of the data resulting from the experiment correctly interpretable. Moreover, efficiency in terms of cost versus precision may be considerably increased by choosing an appropriate design. The aim of this course in not only to teach students to design studies, but also more generally to broaden their understanding of the relation between experimenting and induction.

Survival Analysis
In this course we cover methods for the design and analysis of studies of right censored time to event data, hence `survival' analysis. Assuming non informative censoring, we estimate survival curves, perform hypothesis tests and predict hazards and survival curves from (time-dependent) covariates. Methods are mostly semi-parametric with a focus on the popular Cox proportional hazards model.  We cover martingale residuals for the assessment of assumptions.   We go as far as the analysis of  competing risks data, but stay shy of the full development of recurrent events.  We apply the methods to study for example time to death, to disease recurrence, to ph.d. publication and to bankruptcy:  a range of exciting  and important applications tackled in SAS here.