December 2, 2013 Leave a comment
Some of what I learned this month from Twitter: new preprint server, Google Scholar Library, papers on citations and p-values, and the most networked science conference ever
In what could be a major development in the culture of publishing, a preprint server for biology, BioRxiv, was launched this month. It is based on the long-running arXiv preprint server used by physicists (and increasingly quantitative biologists). Nature News had a good summary.
Google Scholar Library
Google Scholar have launched a new service, Google Scholar Library (h/t @phylogenomics). This is meant to be a way to organize papers you read or cite, so it could be a competitor to reference managers such as Mendeley and Zotero. However, it doesn’t seem to be fully set up for citing papers yet: you can import into BibTeX, EndNote, RefMan and RefWorks (but not Mendeley or Zotero) or get a single citation in just MLA, APA or Chicago style.
“Top researchers” and their citations
- It gives a feeling for what makes a high h-index: of over 15 million authors, about 1% had an h-index of over 20, about 5000 over 50 and only 281 over 80.
- It shows how different sources of citation data can give different h-indices for the same author (see Table 3 in the paper; as pointed out by @girlscientist)
The paper is limited by its reliance on citation data and the h-index alone, so should not be taken too seriously, but it is worth a look if you haven’t already seen it.
p-values vs Bayes factors
The second is a paper in PNAS by Valen Johnson (covered by Erika Check Hayden in Nature News) suggested that the commonly used statistical standard of a p-value less than 0.05 is not good enough – in fact, around a quarter of findings that are significant at that level may be false. This conclusion was reached by developing a method to make the p-value directly comparable with the Bayes factor, which many statisticians prefer. As I’m not a statistician I’m not in a position to comment on the Bayesian/frequentist debate, but it is worth noting that this paper recommends a p-value threshold of less than 0.005 to be really sure of a result. A critical comment by a statistician is here (via @hildabast).
- The keynote talk by Salvatore Mele of CERN. This was not only an accessible explanation of the search for the Higgs Boson, and of the importance of open access and preprint publishing in high energy physics, but also a masterclass in giving an entertaining and informative presentation.
- The discussion session Open, Portable, Decoupled – How should Peer Review change? (Storify of the tweets here)
- The discussion session Altmetrics – The Opportunities and the Challenges (summary and some related links from Martin Fenner here)
- A workshop I helped with, on rewriting scientific text using only the thousand mostly commonly used words in the English language (report by the organiser, Alex Brown, here)