 |
 |
Dear all, On Wednesday 27th October, the RSS Leeds/Bradford local group will be hosting an afternoon of talks on "New developments in visualisation" featuring Roy Ruddle (School of Computing, University of Leeds) and Dean Langan (Clinical Trials Research Unit, University of Leeds). The meeting will be held at the NHS Information Centre, in the Hill and Diggory rooms on the ground floor, starting at 3pm with refreshments from 2.30pm. In accordance with security procedures at the Information Centre, those attending must pre-register. If you plan to attend, please email Paul Baxter ([log in to unmask]) by 5pm on Monday 25th October. On arrival at the Information Centre please go to Reception and ask for the Royal Statistical Society meeting, you will be collected from there. Further details can be found on our webpage: http://tinyurl.com/rss-lba Regards, Paul =================================================================== Dr. Paul D. Baxter Secretary/Treasurer, RSS Leeds/Bradford Local Group, Division of Biostatistics, University of Leeds, Leeds, LS2 9JT, UK. ------------------------------------------------------------------- Leeds/Bradford: Wednesday 27 October, 3.00pm, NHS Information Centre. New developments in visualisation Roy Ruddle (University of Leeds) [Presentation]. Using "giga-pixel" displays to analyse large data sets Large, ultra-high resolution computer displays can be built from commodity hardware, and are colloquially known as "giga-pixel" displays even though they typically have 50+ million pixels rather than a billion. We have built four such displays, two installed on the Leeds University campus and two more at St James' Hospital. In this talk I will explain what we have learned about the benefits of using giga-pixel displays to analyse giant images and spatial datasets, in both a single-user and collaborative (small group) setting. Dean Langan (University of Leeds) [Presentation]. A range of new graphical ideas for meta-analysis My work involves the development of a range of graphical ideas based on the funnel plot. The primary motivation is in light of the recent introduction of numerous additional features which have become widely used such as the contour enhanced funnel plot [1] and the pseudo 95% confidence interval [2]. The popularity of these additional features suggests that the standard funnel plot is currently not being used to its full potential. Such new features all have one aspect in common; they provide a visual illustration of the impact new studies have on a given meta-analysis. Specific features include (1) the statistical significance contour; this feature shows graphically the regions of the funnel plot where one new study would have to be in order to change the statistical significance of the meta-analysis based on a predefined significance level. (2) The heterogeneity contours show how one new study can affect the level of heterogeneity in a given meta-analysis which could lead to switching from fixed to a random effects meta-analysis or vice versa. The two above features and many more have been implemented into meta-analyses indicating powerful and previously unconsidered interpretations are possible. The features within the new graphical framework aim to produce a well rounded graphical display, capable of illustrating the key components of meta-analysis effectively such as the distribution of studies, the level of heterogeneity and presence of bias. From reviewing current graphical displays in meta-analysis, it is clear that a number of current plots would be required outside of the new graphical framework. These features can be used to: 1. Show the current robustness of the meta-analysis. 2. Aide sample size calculations and design prospective similar studies. 3. Help prioritise meta-analyses most in need of an update. References: [1]. Peters et al (2008). Contour-enhanced meta-analysis funnel plots help distinguish publication bias from other causes of asymmetry. Journal of Clinical Epidemiology 61 991-996 [2]. Stern, J et al (2004). Funnel plots in meta-analysis. The Stata Journal 4, Number 2, pp. 127-141 You may leave the list at any time by sending the command SIGNOFF allstat to [log in to unmask], leaving the subject line blank.