From October 2010 to October 2013, I was recipient of a teaching grant.
Amount of hours spent teaching is approximatively 64 hours a year. About one third represents lectures, the remaining consists in tutorials.
List of lectures and tutorials is available below. Only copyright free teaching material is available for download on this page.
Other course materials can eventually be sent on specific request by emailing me (melanie.prague[AT]isped.u-bordeaux2[DOT]fr).
Teaching statement When introducing a new concept, I first introduce the purpose and applied relevance, followed by the intuition
of how the methods work. Depending on the course and the background of students I then go into more mathematical/statistical details.
I also believe students learn better when they are more involved
Lectures and Tutorials (12 hours)
Introduction to Bayesian Statistics philosophy: priors, likelihoods and posteriors. Description of Bayesian inference methods: Point estimation, credibility sets. Application in hierarchical models. Simulations methods with Monte-Carlo Markov Chain samplers: Métropolis-Hastings, Gibbs sampling, simulated annealing sampling... Illustration on linear models and non-linear models with mixed effects. Utilization of Winbugs and R as suitable softwares for Bayesian Statistics.
Tutorials (16 hours)
In this lecture sampling and estimation methods will be tackled. Then, we will focus on tests: adjustment and homogeneity parametric tests (for means, standard deviations and frequencies), chi-deux tests, adjustment non-parametric tests (Smirnov), homogeneity non-parametric tests (Wilcoxon) for independant and matched data, ANalysis Of the VAriance (ANOVA).
Integrated Lectures and Tutorials (6 hours)
SAS is one of the most used statistical software in Pharmaceutical labs and biostatistics engineering. In this integrated tutorial, we introduce what is SAS and how to use it in the Windows environment. We focus on SAS data step to handle data and files types. We illustrate the use of frequently used procedures (FREQ, UNIVARIATE, ...). Basics statistical procedure for estimation and tests in classical models are presented (GLM, NLMIXED, ...). How to make plots is briefly explained (GPLOT, ...).
Description of the curriculum : L1 Social sciences
Tutorials (16 hours)
This consists in giving the basis in statistics and probabilities. The whole statistical vocabulary compulsory to go further in statistics is introduced. We focus on graphical and computational techniques to correctly and efficiently describe a dataset. We tackle linear regression and independence testing. Applied examples are performed with the Excel Software as a statistical tool.
Description of the curriculum : L1/2 Medecine
Integrated lectures in tutorials (20 hours)
Students in medicine have to use office software during their studies. This initial course introduce the use of Word for text edition, Excel for computations and Power Point for slides preparation and presentation techniques (or assimilated free software in OpenOffice or LibreOffice).
Tutorials : Word, Excel, Power Point ...
Description of the curriculum : DU Public health
Correspondence courses by Internet: lectures and tutorials.
The aim of this module is to study the link between a continuous variable and another variable in class (Analysis of Variance) or continuous (correlation, linear regression). Illustrations and computation methods are performed with Excel, EpiInfo and R.
Damien Fossat-Cercler, six months internship, Master’s Degree in Engineering in Statistics, ENSAI, Rennes.
First step of this work is to compare the performances of NIMROD program (written in Fortran, see Current Projects for more details) with other existing software. Pharmacokinetics dataset already used by Plan et al. (2012) will be a good basis for comparison. Vectors of amelioration of the program will be tested such as the change in the methods to account for random effects. Adaptive Gaussian Quadrature will be compared with non-adaptive Gaussian-Quadrature and quasi Monte-Carlo techniques. Finally, more complex illustration will be done with the analysis of observational data from GECSA (HIV infected patients Cohort in Aquitaine). The aim will be to account for many treatment and estimate their effects with dynamical models so as to be able to do pharmacoepidemiology and treatment optimization.
Ana Jarne, six months internship, Master of Science in Biostatistics, Pau Pays de L'Adour University.
Background and aim: To model CD4+ lymphocytes in patients with HIV infection when they receive IL7 by injection and to introduce a feedback factor.
Methods: Investigation of best model was performed on INSPIRE study data, where 27 patients have received recombinant IL7 injections. Estimations are done thanks to the NIMROD program based on penalized likelihood maximization. Model choice was done by comparing LCVa evaluating the data fit performance.
Results: We fitted a system of linear differential equations, with effect of IL7 related to the dose by a pharmacodynamic function. Then, we have designed a non linear model with a feedback factor that fitted the data better. Tentative to use tissue compartment pharmacokinetics concentrations instead of the crude dose of IL7 are in progress during Ana Jarne's PhD continuation with Daniel Commenges.