Sunday, April 19, 2009
One way to avoid scientific criticism
Why isn't it scientific misconduct to delete data in support of a peer-reviewed journal article? Is it because the "science" in question supported, rather than refuted, the dominant paradigm?
If you are going to argue that the output of a computer model represents "science," the least you can do is preserve the data in a useful format so that others can inexpensively replicate the experiment. Otherwise, you are a clown, not a scientist.
Anybody want to take the other side?
4 Comments:
By 
, at
Sun Apr 19, 04:28:00 PM:
Al has stated that the science is settled!!! End of the dialogue. As the libs state, move on to the important issues. Don't ever question The One or Fal Al ever again, racist right wing tactics. Homeland Security is watching you guys.
By Brian Schmidt, at
Mon Apr 20, 04:45:00 PM:
I'm happy to take the other side, TH. The data isn't readily available, but it isn't lost either. Steve just has to go and do what C+B did and generate the derive the data from an existing set.
It's also worth noting that Connolley is no longer at BAS and their data set is likely from pre-2007. Plenty of ways for things to go missing, especially something that's easily recreated if Steve really wants to do it.
By 
TigerHawk, at
Mon Apr 20, 08:53:00 PM:
Brian, I do not know much about science (obviously), but I know a fair amount about auditing. It seems to me that if you are going to make a scientific argument from the results of a model the crunches data, the only way to audit that argument (or the paper) is to run the exact same data through it. Failure to retain that data and make it available for others to review strikes me as sloppy indeed, for it destroys the "chain of custody," if you will, of the information used to generate the results in the paper. Or am I wrong (which I genuinely admit is possible)?
By Brian Schmidt, at
Tue Apr 21, 08:19:00 PM:
I don't want to overstate my amateur science credentials either, but it seems Steve was saying that it takes a lot of work to go from the raw data to the dataset used in the model and he wanted to skip that step by using the original author's work. That's all that's going on, AFAIK.
I've corresponded with Connolley a fair amount in the last few years; I can ask him about his take on all this.
