Skip to content

Dataset size in R

2 messages · Peter.Watkins@foodscience.afisc.csiro.au, Uwe Ligges

#
Peter.Watkins at foodscience.afisc.csiro.au wrote:

            
Well, in theory, there are some limits re. size of some integers, but 
you are far away, as far as I understand.

In practice, you will somewhere hit the limit of available memory in 
your machine.
I don't understand this point:
How many observations and how many variables?
Depends on the problem. For really huge data, you might want to use a 
"64bit machine", 64bit adressing for a single R process is (yet) only 
supported on Linux. If memory does not matter, Linux is sometimes a bit 
faster, but it depends also on other facts like using CPU optimized 
linear algebra systems such as ATLAS or Goto's BLAS.

Uwe Ligges