 |
 |
We are pleased to announce to following half day workshop at the Royal Statistical Society: Classifications and Applications in Big Databases Wednesday December 3rd 2014, 1.00-5.00pm, RSS Errol Street, London. You must register in advance to attend. Lunch is provided. All are welcome. For registration details see: http://www.statslife.org.uk/events/eventdetail/303/14/classification-methods-and-applications-in-big-databases Please note that Daniel Laurison will now speak instead of Mark Taylor. More details about the speakers and their talks are given below. See you there, Pia Hardelid, Katie Harron and Annie Herbert ----------------------------------------------------------- Speakers and abstracts: “Modelling Expert Uncertainty in Medical Decision Support”, Prof. Jon Garibaldi Medical decision making is often difficult, requiring complex decisions in the precence of much uncertainty (both in data and in domain knowledge). In this talk, I shall present some of the recent work we have carried out in modelling uncertainty whilst performing various clustering and classification tasks to support medical decision making. In particular, I will focus on the variability exhibited by human decision makers, and how modelling this may lead to improved decision making. “Multiple Correspondence Analysis and Clustering on Social Network Attributes in the Great British Class Survey”, Dr. Daniel Laurison Abstract to follow. Title TBA, Dr. Emily Petherick Abstract to follow. “Mythbusting with Latent Class Models”, Dr. Matthew Sperrin Latent class models are often used to represent distinct subpopulations within a heterogeneous population. Ecological fallacy can arise when these latent classes are aggregated, leading to erroneous statements being made about a population. We present examples of this in the obesity epidemic, in longitudinal trajectories of disease risk factors, and in the ‘atopic march’ of allergy in childhood. You may leave the list at any time by sending the command SIGNOFF allstat to [log in to unmask], leaving the subject line blank.