automate writing formulas

[Kamil Toth <kamiltoth@yahoo.com>]

Folks:

Many thanks to Arthur Nadas, Michael Camilleri, Frank E Harrell Jr, Patrick Burns, Douglas Bates, Thierry Onkelinx, and Alan Hochberg for help with my request

"automate writing formulas." Most of the responses were variations of the same idea best formulated in the Arthur's message,

as.formula(paste("F~",paste("x",1:10,"+",sep="",collapse=""),"x11",sep=""))

I have also enjoyed a high-minded philosophical dispute which was sparked by my seemingly trifle question. We have learned important things about usefulness of computers, about what to compute, what not to compute, and what damn fools can possibly do if they are not properly schooled on what are computers actually for. I believe that majority of people with philosophical and mentorial tick tried to imagine who the hell this stupid guy is, who is trying to fit linear model with 1000 variables. The best this idea was expressed by Dr. Seneca who said, directly aiming at me, "What fools these mortals be" (Thank you, Alan).

 

Thank you all for useful discussion. For those who was a kind of outraged with my stupidity, I would like to reveal my secret. My problem had nothing to do with statistics. I have translated the problem into statistical language in such a way that a person with statistical background could easily digest it. Simply, I wanted to piggyback some of the "lm" functionality and to create a very fast (and very dirty) solution of a purely computational task. I did not plan to spend more than a couple of hours on this "project", that's why I did not want to go into troubles of FORTRAN programming. Inability to write a lengthy formula was a kind of an unexpected and annoying impediment. Also, the number 1000 has been selected to be a synonym for "many". How many does not matter. Just sooooooooooooooooooooo many that the idea to write it manually would be unacceptable because it is very looooooooooooooooooooooooooong. The question I have asked was not how to do lm-ing. My question was how to write formulas. Thanks again to Arthur who was very fast to answer my question, directly to the point and without attempts to lecture on unrelated topics.

 

By the way, for those with statistical background only: please be aware that optimization in the parameter space of dimension 12,000,000 is not something unusual and utterly outrageous in the modern world. Even in terribly nonlinear problems, even in "assimilation", not just "fitting", mode.

Thanks again for valuable help.

Kamil.

---

Groups are talking. We´re listening. Check out the handy changes to Yahoo! Groups.