| To: | s-news@lists.biostat.wustl.edu |
|---|---|
| Subject: | too big for predict.bdGlm? |
| From: | monika.buholzer@css.ch |
| Date: | Fri, 5 Dec 2008 07:57:43 +0100 |
|
Hi I've installed the bigData library to compute a lineare model for a quite big data set of 1'000'000 x 43 (because I have many factors and interactions I finally get a dataframe of 1'000'000 x 246). The linear model I do compute for a small sample of 200'000 x 43 and it works without any problems. But when I want to predict the values for the whole data set of 1'000'000 x 43 S-Plus tells me, it hasn't enough dynamic memory. Is there a chance to get along with this problem? I now have to predict five times for a sample of the data. If I gonna have bigger data sets in the future, it will become quite boring... Thanks a lot Monika ___________________________________________________________________ CSS Versicherung Monika Buholzer FSST Mathematikerin Tribschenstrasse 21 CH-6002 Luzern Telefon ++41 (0)58 277 18 11 Telefax ++41 (0)58 277 12 12 monika.buholzer@css.ch www.css.ch ___________________________________________________________________ Unternehmen der CSS: CSS Kranken-Versicherung AG, CSS Versicherung AG, Arcosana AG, Auxilia Kranken-Versicherung AG, vivit gesundheits ag
------ |
| <Prev in Thread] | Current Thread | [Next in Thread> |
|---|---|---|
| ||
| Previous by Date: | Re: sample command as FUN element using tapply, Tim Hesterberg |
|---|---|
| Next by Date: | Announcing TIBCO Spotfire S+ 8.1, Louis Bajuk-Yorgan |
| Previous by Thread: | Jan-Feb Schedule: Interactive Online Statistics Courses, courses |
| Next by Thread: | Announcing TIBCO Spotfire S+ 8.1, Louis Bajuk-Yorgan |
| Indexes: | [Date] [Thread] [Top] [All Lists] |