Hello, I have 2D-data on a torus, i.e. they are scattered within [0:2pi) and are supposed to be periodic with period 2pi. Is there a way in R for a kernel density estimation for such data? I found this article http://www.dmqte.unich.it/personal/dimarzio/density46.pdf but a) I don't fully understand the article (my knowledge in statistics is poor) b) I did not understand which Eq. represents the kernel(s) c) I do not now R well enough to understand whether I can use kde2d or nprudens with an arbitrary kernel My simple-minded attempt was to extend the data to [-2pi:4pi) and then use kde2d, but I am wondering a) how accurate is this b) is there a way to do it properly on a torus. Thanks a lot, Tim -- -- Tim Gruene Institut fuer anorganische Chemie Tammannstr. 4 D-37077 Goettingen GPG Key ID = A46BEE1A -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 189 bytes Desc: Digital signature URL: <https://stat.ethz.ch/pipermail/r-help/attachments/20100728/be5aebc7/attachment.bin>
On Jul 28, 2010, at 7:55 AM, Tim Gruene wrote:> Hello, > > I have 2D-data on a torus, i.e. they are scattered within [0:2pi) > and are > supposed to be periodic with period 2pi. > > Is there a way in R for a kernel density estimation for such data? I > found this > article http://www.dmqte.unich.it/personal/dimarzio/density46.pdf > but > a) I don't fully understand the article (my knowledge in statistics > is poor) > b) I did not understand which Eq. represents the kernel(s) > c) I do not now R well enough to understand whether I can use kde2d > or nprudens > with an arbitrary kernel > > My simple-minded attempt was to extend the data to [-2pi:4pi) and > then use > kde2d, but I am wondering > a) how accurate is this > b) is there a way to do it properly on a torus.You could take a look at the density function within the package, "circular". -- David Winsemius, MD West Hartford, CT