The page says the following: "Log file analysis: Andes raw logs can be converted to the DataShop format, but the converted logs do not have the right information in them for the kinds of analysis experimenters want to do."
It would be good to have an example of an analysis that an experimenter wants to do that cannot currently be done in the DataShop.
One prior concern was that the DataShop's use of the "location heuristic" for error attribution would not work well for Andes data and experimenters needs. It need be true for some needs (it would be good to articulate those), but for the purposes of generating "smooth" learning curves, the analysis done in the following paper suggested that the location heuristic does as well or better than alternatives.
Nwaigwe, Adaeze; Koedinger, Kenneth; VanLehn, Kurt; Hausmann, Robert, Weinstein Anders. (in press). Exploring Alternative Methods for Error Attribution in Learning Curves Analysis in Intelligent Tutoring Systems. To appear in the proceedings of the International Conference on Artificial Intelligence in Education 2007.
There may be other concerns besides this one. It would be good to articulate them.
(I'm not sure what the right procedure is for using the talk pages!?)
Ken, That was a leftover from 6 months ago. In the last few weeks, we have finally gotten some real learning curves out. I will make the update. Brett