Re: help: testing for unimodality in multiple dimensions?
- From: hrubin@xxxxxxxxxxxxxxxxxxxx (Herman Rubin)
- Date: 28 Dec 2005 13:38:59 -0500
In article <1135783712.234745.140830@xxxxxxxxxxxxxxxxxxxxxxxxxxxx>,
Eric B <eblabac@xxxxxxxxxxxxxxxxxxx> wrote:
>You might want to consider looking more closely at the contour plots
>... you might be able to tell just by looking at it whether or not its
>unimodal. Also to check for unimodality you could use the
>Kolmogorov-Smirnov (KS) test in R, since you have your data (call it
>x), you could generate any 5-d distribution in R with one mode (say a
>multivariate normal), and use the KS test. The KS test is a two-sample
>test of the null hypothesis that 'x' and 'y' were drawn from the same
>continuous distribution.
>Hope that helps.
>Eric B
Even in one dimension, the distribution of the K-S test
statistic changes if any parameters are estimated.
In more than one dimension, the distribution of the
K-S statistic is not know, and depends on the
multivariate distributions, even if consistent.
--
This address is for information only. I do not claim that these views
are those of the Statistics Department or of Purdue University.
Herman Rubin, Department of Statistics, Purdue University
hrubin@xxxxxxxxxxxxxxx Phone: (765)494-6054 FAX: (765)494-0558
.
- References:
- help: testing for unimodality in multiple dimensions?
- From: raacampbell
- Re: help: testing for unimodality in multiple dimensions?
- From: Eric B
- help: testing for unimodality in multiple dimensions?
- Prev by Date: Re: Mann Whitney U test
- Next by Date: Factor analysis: Computing factor scores
- Previous by thread: Re: help: testing for unimodality in multiple dimensions?
- Next by thread: Factor analysis: Computing factor scores
- Index(es):
Relevant Pages
|
