Hello, I am quite new in R, and I have one problem: I have large d-dimensional data sets (d=2, 3, 6, 10). I would like to divide the d-dim space into n (n may be 10, but better some larger number, for example 20) equally sized d-dim hypercubes and count how many data points are in each cube. Is there any way to do it quickly, I mean - in a reasonable time? Actually, I want to get some rough idea of underlying densities of these data and compare them. Thanks a lot! Anna
Quick partitioning
2 messages · Anna Oganyan, Brian Ripley
On Thu, 25 Aug 2005, Anna Oganyan wrote:
Hello, I am quite new in R, and I have one problem: I have large d-dimensional data sets (d=2, 3, 6, 10). I would like to divide the d-dim space into n (n may be 10, but better some larger number, for example 20) equally sized d-dim hypercubes and count how many data points are in each cube. Is there any way to do it quickly, I mean - in a reasonable time? Actually, I want to get some rough idea of underlying densities of these data and compare them. Thanks a lot! Anna
How do you divide a 10D space into 10 hypercubes? You need at least some of dimensions to be undivided. The general idea is easy: apply cut() to each dimension, so your dimensions become factors, then table() will produce the counts. That will be quick enough for millions of points.
Brian D. Ripley, ripley at stats.ox.ac.uk Professor of Applied Statistics, http://www.stats.ox.ac.uk/~ripley/ University of Oxford, Tel: +44 1865 272861 (self) 1 South Parks Road, +44 1865 272866 (PA) Oxford OX1 3TG, UK Fax: +44 1865 272595