LEUVEN STATISTICS RESEARCH CENTRE (LSTAT) CELESTIJNENLAAN 200 B BOX 5307 3001 LEUVEN, BELGIË Thesis topic (before March 13, 2014) Title : A comparison of robust correlation estimators OUR REFERENCE YOUR REFERENCE LEUVEN Name promoter: Stefan Van Aelst Available for students from - Biometrics, General Statistical Methodology Description: Indicate whether the working area is the KULeuven or a company outside KULeuven. The motivating problem is the need to calculate many correlation measures to unravel the relation between the major RNA parts (mRNA) and more recently explored pieces of RNA such as microRNA (miRNA). To understand the role of these miRNA pieces, correlations are examined with many wellknown mRNA parts. Newer bio-technology tools are cheaper and faster than before, but also more error-prone, so the resulting data may be contaminated. Given the large size of such data (hundreds to thousands of RNA pieces), it is not feasible to ‘clean’ the data by hand. Therefore, the correlation measures calculated from the data need to be robust. Moreover, they need to be easy to compute because several millions of correlations need to be calculated. The purpose of this thesis is to investigate properties of existing and new correlation measures, such as their robustness and accuracy, but also their computation time will be examined. The comparison will involve well-known correlation measures such as Spearman and Kendall correlation, but also correlation measures based on robust orthogonal regression estimators. For the application the ranking of RNA pieces according to their correlation is of great importance, so it will be investigated to what extent the rankings corresponding to the different correlation measures differ from each other. The thesis work will be carried out at KU Leuven. AN CARBONEZ TEL. + 32 16 32 22 42 An.Carbonez@lstat.kuleuven.be lstat.kuleuven.be