Apologies Phil if you felt that these comments were personal in any way; this was not my intent. My aim was to correct errors in The Scientist's coverage and alert readers to possible conflicts of interest that are not immediately apparent in the primary article or The Scientist's piece.
Regarding the availability of the dataset in this study, it is a fact that it was not made available at the time of publication. It is true that I did not make our private correspondence public, since I would have assumed you would view this as a breach of confidence without first asking for your permission to quote you. But as you have chosen to make this exchange public, I will note for the record that you refused to provide a dataset that included DOIs of the papers in this study, because of restrictions on data sharing imposed by the publishers (some of whom are also funders) that gave you access to data. The reason this restriction is important is that no scientist can match download statistics in your anonymized dataset back to the original papers to investigate underlying biases and confounding factors in the "control" and "PMC" sets (e.g. research topic), which may explain your results.
This refusal to make data available for extension and deeper investigation, not merely statistical replication, is of course not a requirement of FASEB J or other journals. Many scientists do not share their data openly, especially when funded by private industry. What is of concern in this case is that the funder of the study, the APS, is also the parent organization of one of the main publishers that both provided data and imposed data sharing restrictions. Moreover, the funders and data providers are also members of FASEB, which publishes the journal in which this study was reported. Given that the main message of the study is that publishers are being "hurt" by PubMed Central, there appear to be several conflicts of interest here regarding society publishers funding, providing data, preventing this data from being openly investigated, and ultimately publishing the study at hand.
Let's be clear here: we are not dealing with sensitive medical or other personal data here, we are dealing with download statistics from journal articles, information that many journals actively give away for free on their websites. I can see no reason that download statistics linked to DOIs cannot be made available for further analysis, other than to impede other scientists from investigating the claims of this study more fully.
Let me also clarify that I have no issue with privately-funded research and research conducted at non-academic institutions being published in academic journals. I fully agree that everyone has a right to practice science, not just those in the ivory tower. My concern here has nothing to do with you affiliation. It has everything to do with real and/or apparent conflicts of interest. I would have the same concerns about any study that was funded by a by a private entity using data provided by that entity, which was not made publicly available for scrutiny, regardless of their affiliation.
Finally, regarding your declaration of possible conflict of interests, according to the FASEB J's guideline (http://www.fasebj.org/site/misc/edpolicies.xhtml#conflict), authors are required to explicitly disclose conflicts of interest in the acknowledgements: "If the article is accepted for publication, information on the potential conflict of interest will be included in the acknowledgments section of the article." Unfortunately, the potential conflicts of interest in this case were not made explicit, and your disclaimer in the paper excerpted above only relates to responsibility of the data/analysis, which is a minimal standard expected for all scientists and does not reveal the nature of the issues at hand. A full disclosure of potential conflicts of interest in this case would have stated something like: "This study was funded by the APS, who provided access to data in the study, and is a member of FASEB who publishes the FASEB J." This sort of clear, direct conflict of interest statement would be an absolute requirement, for example, in a study on the effects of a drug funded by a pharmaceutical company using data from a private research lab. I see no difference in this case.