Projects proposition
David Chavalarias - visualisation of social cognitionI would like to propose a session about a project aiming at monitoring and visualize science dynamics from electronique publications and more broadly social dynamics from online social media. The work will require to built crawlers to retreive online data and the program to treat and visualize them. The session will begin by a branstorming about these issue to reach some concrete feasible ideas. The rest of the session will be for the implementation. Ressources : Paper are available online on my Website as well as the website of the MOMA group (http://moma.csregistry.org).
John Ioannidis and Thomas Pfeiffer: The spread of false findings in networks or scientists
The spread of findings in the scientific community might be modeled as an epidemic where instead of a virus, scientific results are spreading. These results can be either true or false. Negative and positive findings, as well as true and false ones may differ in their infectivity and other parameters. The dynamics of knowledge in such a network can be modeled for different distributions of contact among scientific articles or scientists; and it can be extended across fields with different levels of cross-communication. As potential outcomes, we can estimate under what conditions false findings reach a critical level, e.g. at least x% of all papers/scientists are infected (use/cite/support claims based on) with at least 1 false-finding or at least y% are infected with x false-findings. We can also see which parameters are most influential in reducing the epidemic in the field – the number and rate of false findings, distribution of infectivity/basic reproductive numbers, clearance rates, etc. John Ioannidis and Thomas Pfeiffer: The Proteus phenomenon and exaggerated effects
In the scientific literature, the description of quantitative effects, such as the influence of a mutation on the chance of developing a certain disease, often follows remarkable patterns. Early findings tend to be the most extreme and often contradict each other (Ioannidis and Trikalinos, 2005). After a few publications, estimates of an effect seem to reach a consensus. Moreover, published effect sizes are often exaggerated (Ioannidis 2008). In the proposed project, we would like to study how these patterns emerge. Is this due to irrational behavior of scientists and publishers, or is there some rational behind these effects? How have the rule of science to be designed to ensure that a reliable consensus if established within short time? These questions can be studied using computer simulations of simple “toy science” scenarios, as well as behavioral experiments (Pfeiffer et al. 2008). J. Ioannidis, T. Trikalinos. Early extreme contradictory estimates may appear in published research: The Proteus phenomenon in molecular genetics research and randomized trials. Journal of Clinical Epidemiology, vol. 58, no. 6, pp. 543–549, 2005 J. Ioannidis. Why Most Discovered True Associations Are Inflated. Epidemiology. in press, 2008 T. Pfeiffer, D.G. Rand, A. Dreber. Decision-making in research tasks with sequential testing. submitted, 2008 pdf's are available in the file archive: http://medilsws.csregistry.org/tiki-list_file_gallery.php?galleryId=1
Anna Dreber - Biases in gender studiesP-values in scientific publications are often biased, because studies that achieve formal statistical significance might be favoured for publication. A way to detect such a bias is to compare p-values reported in studies that were explicitly looking at gender differences with p-values from studies that focused on different effects but report a gender effect because the corresponding data were collected anyway. I propose a project to study publication biases by comparing p-values from these two types of studies.
|
Except where otherwise noted, content on this site is
licensed under a Creative Commons Attribution 2.5 License
Wiki
Calendars