[R] quantile regression: out of memory error

Prew, Paul Paul.Prew at ecolab.com
Mon Jul 11 19:39:54 CEST 2011


Hello,  I’m wondering if anyone can offer advice on the out-of-memory error I’m getting. I’m using R2.12.2 on Windows XP, Platform: i386-pc-mingw32/i386 (32-bit).

I am using the quantreg package,  trying to perform a quantile regression on a dataframe that has 11,254 rows and 5 columns.

> object.size(subsetAudit.dat)
450832 bytes

> str(subsetAudit.dat)
'data.frame':   11253 obs. of  5 variables:
 $ Satisfaction     : num  0.64 0.87 0.78 0.75 0.83 0.75 0.74 0.8 0.89 0.78 ...
 $ Return           : num  0.84 0.92 0.91 0.89 0.95 0.81 0.9 0.87 0.95 0.88 ...
 $ Recommend        : num  0.53 0.64 0.58 0.58 0.62 0.6 0.56 0.7 0.64 0.65 ...
 $ Cust.Clean       : num  0.75 0.85 0.72 0.72 0.81 0.79 0.79 0.8 0.78 0.75 ...
 $ CleanScore.Cycle1: num  96.7 83.3 93.3 86.7 96.7 96.7 90 80 81.7 86.7 ...

rq(subsetAudit.dat$Satisfaction ~ subsetAudit.dat$CleanScore.Cycle1, tau = -1)

ERROR:  cannot allocate vector of size 2.8 Gb

I don’t know much about computers – software, hardware, algorithms – but does this mean that the quantreg  package is creating some massive intermediate vector as it performs the rq function?   Quantile regression is something I’m just starting to explore, but believe it involves ordering data prior to the regression, which could be a huge job when using 11,000 records.   Does bigmemory have functionality to help me with this?

Thank you,
Paul






Paul Prew   ▪  Statistician
651-795-5942   ▪   fax 651-204-7504
Ecolab Research Center   ▪  Mail Stop ESC-F4412-A
655 Lone Oak Drive   ▪   Eagan, MN 55121-1560




CONFIDENTIALITY NOTICE: 
This e-mail communication and any attachments may contain proprietary and privileged information for the use of the designated recipients named above. 
Any unauthorized review, use, disclosure or distribution is prohibited. 
If you are not the intended recipient, please contact the sender by reply e-mail and destroy all copies of the original message.



More information about the R-help mailing list