If you can get model.matrix to make an X matrix for you, then you could try to use slm in SparseM to estimate the model. Once again, I would make a plea that it would be nice to have a version of model.matrix that returned a sparse form of X, since there are bound to be problems for which X itself creates memory problems even before one tries to hit it with the QR hammer.
[SNIP]
Dear all, before this thread grows further on:
Silika told me in a private message of the size of the problem. When I guess that a simple additive linear model should be applied (I still don't know), it is solvable by lm() with less or than 512MB RAM (tested on my machine), but not with 128MB on his/her Laptop ...
Thus, resizing the amount of RAM for R might solve the problem in a reasonable amount of time given not too much swapping occurs.
Best, Uwe
______________________________________________ [EMAIL PROTECTED] mailing list https://www.stat.math.ethz.ch/mailman/listinfo/r-help
