PLoS ALM on OpenSci

Has anyone generated a slightly nicer data object out of this data yet? Been thinking of graphing the correlations of downloads versus citations versus whatever and similar for different journals which really requires a bit of cleaning up the data to be effective but if someone has already done it? - Cameron Neylon

Cameron: what else needs to be done to make the data more usable? the source data here is available as TSV http://manyeyes.alphaworks.ibm.com/manyeye... and CSV or XLS too, is there any other format that would be better? - Mike Chelen

I was wanting to do some analysis that included comparing papers based on time of publication i.e. "what is the average trajectory of downloads?" as well as comparing these across journals so I was hoping someone might have converted to either SQL and/or a set of python objects containing lists of downloads/citations/pageviews by month. Not difficult to do myself but just wondered whether someone else had already. - Cameron Neylon

is this other ManyEyes dataset helpful? http://manyeyes.alphaworks.ibm.com/manyeye... it contains the "per day" figures for most of the metrics, including PDF and XML downloads. some spreadsheet and CSV files are also in a github repository http://github.com/mchelen... which might be convenient for import to SQL - Mike Chelen

That would certainly do one of the things I had in mind but the big problem I was having was with wanting to come up with average initial rates and saturation points to see if there are any characteristics of "hot" vs "slow-burn" papers. I saw some evidence of this in the very crude graph analysis I did when the stats first came out. - Cameron Neylon from twhirl

how could the change in rates for each article be determined given only the totals? while the plos website includes a chart of an article's recent history, the data released so far can show how older and newer articles compare in terms of downloads per day like PDF files http://friendfeed.com/mikeche... - Mike Chelen

Cameron: here can be seen which years and journals have articles with the most downloads per day http://friendfeed.com/mikeche... is this close to what you have in mind? - Mike Chelen

Looks very nice Mike. We should have all the 'missing' usage data (pre Aug 2005 and first 200 PLoS ONE articles) added sometime this week. - Peter Binfield

Peter: great thanks, looking forward to it! any preferences or suggestions about where people might want to look for or share data analysis results? - Mike Chelen

Mike, just realised that I've got a somewhat different dataset that I think hasn't been publicly released yet which includes all these parameters by month - but as Pete points out there are some dates missing. - Cameron Neylon

Actually Cameron, you dont. We have released the usage data down to the month level (and you may be referring to that), but not the citation/bookmarks/blogs/comment/notes etc data (although we track cumulative data on these items, by the month, we only started tracking it in March, so dont really have enough monthly data to release - though we could if people felt it was valuable) - Peter Binfield

Cameron: how about this to see differences in citation source (scopus, pubmed, crossref) grouped by journal: http://friendfeed.com/mikeche... or similarly to look at the combined citation types of each journal: http://friendfeed.com/mikeche... - Mike Chelen

Peter & Cameron: ah yes, there is history data is in the other sheets, hadn't even started looking past the first one =) - Mike Chelen