R help - Jul 2008 - Quantitative analysis of non-standard scatter plots.

Hi,

I am having difficulties in finding ways to analyse scatter plots and
quantitatively differentiate between them. Since this is best demonstrated
by a figure, I have attached one with 4 subfigures (a)-(d). Qualitatively
(and I hope you would agree with me), sub-figures (a), (b), and (d) seem to
represent uniform 2d scatters. It is hard for me to quantify it, but it
seems as if the envelopes (max values of y-axis) are independent of the
values of the x-axis. Sub-figure (c), on the other hand, exhibits a
different behaviour, as the envelope seems to expand as the x-values
increase (at least till some threshold, and with the exception of a single
outlier toward the begining of the x-axis). My question is how can I verify
the above observation quantitatively, and does R provide the required tools
for doing so? I have tried to perform linear fittings and to claim that the
slope of the fitted line in (c) is more significant than the lines in (a),
(b), and (d), but this does not seem convincing to me. Are there standard
toolboxes for such cases?

Thanks for any suggestions / tips,
Firas.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: scatterPlots.png
Type: image/png
Size: 83291 bytes
Desc: not available
URL:
<https://stat.ethz.ch/pipermail/r-help/attachments/20080723/3ab73a81/attachment.png>

Firas Swidan <frsswdn <at> gmail.com> writes:
> 
> Hi,
> 
> I am having difficulties in finding ways to analyse scatter plots and
> quantitatively differentiate between them.  
  Try quantile regression, implemented in Roger Koenker's quantreg
package.  There's a very thorough vignette (library(quantreg);
vignette("rq")).

  Ben Bolker

Have u checked skewness of data?

-----Original Message-----
From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-project.org]
On
Behalf Of Firas Swidan
Sent: 23 July 2008 18:38
To: r-help at r-project.org
Subject: [R] Quantitative analysis of non-standard scatter plots.

Hi,

I am having difficulties in finding ways to analyse scatter plots and
quantitatively differentiate between them. Since this is best demonstrated
by a figure, I have attached one with 4 subfigures (a)-(d). Qualitatively
(and I hope you would agree with me), sub-figures (a), (b), and (d) seem to
represent uniform 2d scatters. It is hard for me to quantify it, but it
seems as if the envelopes (max values of y-axis) are independent of the
values of the x-axis. Sub-figure (c), on the other hand, exhibits a
different behaviour, as the envelope seems to expand as the x-values
increase (at least till some threshold, and with the exception of a single
outlier toward the begining of the x-axis). My question is how can I verify
the above observation quantitatively, and does R provide the required tools
for doing so? I have tried to perform linear fittings and to claim that the
slope of the fitted line in (c) is more significant than the lines in (a),
(b), and (d), but this does not seem convincing to me. Are there standard
toolboxes for such cases?

Thanks for any suggestions / tips,
Firas.

How do I check that, and what quantification does it provide?

Thanks,
Firas.
On Wed, 2008-07-23 at 21:20 +0530, Bogaso Cristofer
wrote:
> Have u checked skewness of data?
> 
> -----Original Message-----
> From: r-help-bounces at r-project.org [mailto:r-help-bounces at
r-project.org] On
> Behalf Of Firas Swidan
> Sent: 23 July 2008 18:38
> To: r-help at r-project.org
> Subject: [R] Quantitative analysis of non-standard scatter plots.
> 
> Hi,
> 
> I am having difficulties in finding ways to analyse scatter plots and
> quantitatively differentiate between them. Since this is best demonstrated
> by a figure, I have attached one with 4 subfigures (a)-(d). Qualitatively
> (and I hope you would agree with me), sub-figures (a), (b), and (d) seem to
> represent uniform 2d scatters. It is hard for me to quantify it, but it
> seems as if the envelopes (max values of y-axis) are independent of the
> values of the x-axis. Sub-figure (c), on the other hand, exhibits a
> different behaviour, as the envelope seems to expand as the x-values
> increase (at least till some threshold, and with the exception of a single
> outlier toward the begining of the x-axis). My question is how can I verify
> the above observation quantitatively, and does R provide the required tools
> for doing so? I have tried to perform linear fittings and to claim that the
> slope of the fitted line in (c) is more significant than the lines in (a),
> (b), and (d), but this does not seem convincing to me. Are there standard
> toolboxes for such cases?
> 
> Thanks for any suggestions / tips,
> Firas.
> 
-- 
Firas Swidan, PhD
Founder and CEO
Olymons: Blessing Machines with Vision (TM)
http://www.olymons.com
P.O.Box 8125
Nazareth 16480
Israel
Cell: +.972.(0)54.733.1788