home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: sci.math.stat
- Path: sparky!uunet!news.claremont.edu!ucivax!ucla-cs!ucla-mic!agsm!iwelch
- From: iwelch@agsm.ucla.edu (Ivo Welch)
- Subject: S: lm() problem.
- Message-ID: <1993Jan12.102240.12909@mic.ucla.edu>
- Nntp-Posting-Host: risc.agsm.ucla.edu
- Organization: UCLA, Anderson Graduate School Of Management
- Date: 12 Jan 93 10:22:40 PST
- Lines: 26
-
-
- [S on NeXT; an April 1992 version; NeXTStep 3.0; 32MB RAM; ~40-80MB free disk]
-
- I am having a serious problem with lm()'s use of disk space. I have a panel
- data set with 1,109 observations and 37 variables, or about 164K in float
- storage. One of these variables is a categorical factor, taking about 400
- values (firm id's). A rough computation tells me that I would need a maximum
- of 1MB (double) space to store one copy of the X'X matrix in full form.
-
- > result <- lm(formula.complete, can, subset= (date<breakdate), singular.ok=TRUE)
- Error in data[1:ll] <- old: Cannot allocate 39516128 bytes:
- options("object.size") is 20000000
-
- Now this is a bit too much for my somewhat limited computer. (After I allow
- more 40MB as object.size, I get a swapspace overflow.) How can lm() be so
- voracious? Is there any way to tell it to be a bit more graceful about its
- memory appetite? (I need to move to about 1,000 variables and 2,000 obs at
- some point.) Obviously, rewriting lm() (for an S beginner like myself) is very
- undesirable.
-
- Please respond to me. I will summarize any helpful suggestions.
-
- /ivo
-
- PS: This is why I cannot do my problem in S. I will post a question on SAS,
- where I also tried to solve my problem soon.
-