R help - Dec 2011 - Statistical tests and measures for "cone-like" distributions?

Hi,

Are there any special statistical tests, or functions in R I could use
to measure "cone-like" distributions?

I have several data-sets, which I've been plotting parts of as 2D
plots,  where I get a "cone-like" distribution of the data points.
That is, the data appears to be bounded by two non-parallel lines,
starting at the origin, giving rise to a "cone-like" appearance.

I guess I could work out it's bounding lines fairly easily, but this
seems perhaps a fairly naive thing to do, and I was wondering if there
are any standard tests to do? I guess one might consider the angle of
the cone, how much of the "cone" is filled out, density of points
within the cone and so on.

I don't even know if thinking of a "cone" makes any sense, what it
would mean or whether I should simply be thinking of a linear
regression line with some non-constant variance?

An example plot is at
http://cs.swan.ac.uk/~csmg/aes/plots/canon_1_3_4_r1_v_cfs.pdf if
anyone is interested.

Thanks in advance!

Matthew Gwynne
http://cs.swan.ac.uk/~csmg/

Have you tried plotting log(y) vs log(x)?

albyn

On Wed, Dec 21, 2011 at 02:18:56PM +0000, Matthew Gwynne
wrote:
> Hi,
> 
> Are there any special statistical tests, or functions in R I could use
> to measure "cone-like" distributions?
> 
> I have several data-sets, which I've been plotting parts of as 2D
> plots,  where I get a "cone-like" distribution of the data
points.
> That is, the data appears to be bounded by two non-parallel lines,
> starting at the origin, giving rise to a "cone-like" appearance.
> 
> I guess I could work out it's bounding lines fairly easily, but this
> seems perhaps a fairly naive thing to do, and I was wondering if there
> are any standard tests to do? I guess one might consider the angle of
> the cone, how much of the "cone" is filled out, density of points
> within the cone and so on.
> 
> I don't even know if thinking of a "cone" makes any sense,
what it
> would mean or whether I should simply be thinking of a linear
> regression line with some non-constant variance?
> 
> An example plot is at
> http://cs.swan.ac.uk/~csmg/aes/plots/canon_1_3_4_r1_v_cfs.pdf if
> anyone is interested.
> 
> Thanks in advance!
> 
> Matthew Gwynne
> http://cs.swan.ac.uk/~csmg/
> 
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
> 
-- 
Albyn Jones
Reed College
jones at reed.edu

Matthew Gwynne <mathew.gwynne <at> gmail.com> writes:
> 
> Hi,
> 
> Are there any special statistical tests, or functions in R I could use
> to measure "cone-like" distributions?
> 
> I have several data-sets, which I've been plotting parts of as 2D
> plots,  where I get a "cone-like" distribution of the data
points.
> That is, the data appears to be bounded by two non-parallel lines,
> starting at the origin, giving rise to a "cone-like" appearance.
  You may be interested in quantile regression.  Roger Koenker, who
has developed a lot of the basic methodology for quantile regression,
is also the author of the quantreg package (which comes with a fairly
thorough vignette IIRC).