It was interesting to hear different opinions about that - Jean-Claude Bradley

Really? Who throws data away? - Bill Hooker

It involved a discussion with Steven Bachrach who made some good points. He was talking about data that he felt utterly useless, such as when a heating switch was just plain not turned on so data were clearly nonsense. (The equivalent in my lab would be if we forgot to turn on the arc lamp and all the images were just black and / or dead pixels.) J-C and I were more of the camp that even stuff we do wrong could _possibly_ be useful and I still believe that. Steve B. did convince me, though that there's a spectrum and some point where it's useless and will vary from lab to lab. More than that, though, his good point was that we need to sell ONS and Open Data to some pretty strong non-believers. So, THOSE people are going to think that most data IS useless. So as far as sales goes, we should not be advertising that all peeps should keep and share everything. Rather, it may be an easier sale to start with ALL data that relates to a given publication (and thus is manifestly "useful"). - Steve Koch

I don't think I agree... warning, half-baked thoughts: ONS is not something that can be sold, for ONS to get any traction requires a wholesale change in the attitude of the working scientist. As you and I have discussed, grad students seem to start out with the ONS attitude and get it beaten out of them. So, realpolitik arguments like SB's don't carry much weight. In fact, a more radical approach ("open means open, no weaseling") might be more attractive to the sorts of people likely to enter the ONS fold. Moreover, one of the benefits of ONS is to historiography of science: how much more profound could the insights of Popper/Kuhn/Lakatos et al have been, had they had access to a *complete* record? And when I say complete, I mean including the ones where the switch was just not turned on. That dataset may not tell us anything about the experimental system but it does tell us something about the people doing the work. So on the whole, I don't really buy the spectrum argument. A notebook that gets groomed before exposure is not Open. I'm not making a value judgement there, it's just not what I think of when I think of ONS. - Bill Hooker

Bill - you expressed my position nicely - ONS gives us an opportunity to spy on how science actually gets done - so we have some hard numbers about how many attempts it takes to get an experiment right, how many times errors are corrected and the timeline of those corrections, etc. An interesting stat is 13% of our reactions are "successful" - meaning they have an isolated yield and characterization. Don Pellegrino is an example of a researcher who is currently mining the ONS metadata we've collected. - Jean-Claude Bradley

I agree with both of you and I also think that knowing about utter failures is important and interesting. Hopefully SB will see this thread and pipe in some. To bring in another comment, Antony Williams also points out the "who's going to pay for it?" question. To see those people, a limited data set may still be necessary. And Bill, I still do think that ONS is natural for uncorrupted scientists. So, maybe if funding agencies require some ONS for even partial work, it will wake up the innate ONS desires and get those peeps to do full ONS. - Steve Koch

throw data away? do these people have any idea how much it costed them to get it?? - Egon Willighagen