Confidence ellipse for simple correlations

[Claudia]

Dear all,

I have been trying to fit a 95% (or 99%) confidence ellipse in simple
scatterplots, but SPSS seems unable to do it. Does anyone know a macro
that can do this?

Thank you in advance for any help,
Claudia

[Chuck Cleland]

Claudia wrote:
> I have been trying to fit a 95% (or 99%) confidence ellipse in simple
> scatterplots, but SPSS seems unable to do it. Does anyone know a macro
> that can do this?

Claudia:
Your request is not entirely clear and you don't say which version of SPSS you
are using, but if you want a CI around the least squares regression line in a 2D
scatter plot, you can do it using interactive graph in version 10.

IGRAPH
/VIEWNAME='Scatterplot'
/X1 = VAR(myXvar) TYPE = SCALE
/Y = VAR(myYvar) TYPE = SCALE
/COORDINATE = VERTICAL
/FITLINE METHOD = REGRESSION LINEAR INTERVAL(95.0) = MEAN
LINE = TOTAL SPIKE=OFF
/X1LENGTH=3.0
/YLENGTH=3.0
/X2LENGTH=3.0
/CHARTLOOK='NONE'
/SCATTER COINCIDENT = NONE.

HTH,

Chuck

-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-
Chuck Cleland
Institute for the Study of Child Development
UMDNJ--Robert Wood Johnson Medical School
97 Paterson Street
New Brunswick, NJ 08903
phone: (732) 235-7699
fax: (732) 235-6189
http://www2.umdnj.edu/iscdweb/
http://members.nbci.com/cmcleland/
-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-<>-

[Claudia]

Hello Chuck,

You are right about my message being unclear. What I actually had in mind was
bivariate outliers. The 95% confidence ellipse, then, would represent the
confidence interval of a bivariate normal distribution (not OLS regression), which
would be superimposed on the scatterplot. Any observations falling outside the
ellipse would then be considered as outliers. This procedure can be easily
implemented in STATISTICA, but apparently not in SPSS (v. 10, by the way), which is
why I am looking for a macro.

Thank you for your help and apologies for being unclear,
Claudia

[Mesbah]

Hi Claudia,
SPSS offers far more flexibility in calculating diagnostics
by allowing you to produce all partial plots, as well as saving
standardized predicted values, Cook's distance, Leverage values,
studentized deleted residuals, standarized DfBetas, and DfFit
values for specified CI, and then use scatterplots to view 2D, 3D
views in which the outliers are immediately.
SPSS 10 3D graphics gives options in defining a nice colorful
presentation of 2-variable CI surface.

Mesbah.

"Claudia" <claudi...@planetaccess.com> wrote in message
news:3ABB762D...@planetaccess.com...

[Fearless]

Use Systat, S-Plus, or R. Systat does this from a menu (as well as in
code), and the code to write a routine in S-Plus or R is relatively trivial.
If you move to R, let me know. I've already written a simple function to
calculate and draw the bivariate confidence ellipses at any arbitrary
p-value.

"Claudia" <claudi...@planetaccess.com> wrote in message
news:3ABB762D...@planetaccess.com...

[Fearless]

This isn't the appropriate place to post non-SPSS code. I'd be happy to
email the crude R functions (plotell() and ellipse() - called by plotell())
to anyone who needs them.

"David Lindsay" <on...@use.reply-to.co.uk> wrote in message
news:O73qhAA$fCz6...@clara.co.uk...
> X-no-archive: yes
> In article <9ag8s...@enews1.newsguy.com>, Fearless <no...@no.way>
> writes

> >Use Systat, S-Plus, or R. Systat does this from a menu (as well as in
> >code), and the code to write a routine in S-Plus or R is relatively
trivial.
> >If you move to R, let me know. I've already written a simple function to
> >calculate and draw the bivariate confidence ellipses at any arbitrary
> >p-value.
> >

> Please post it sounds interesting.
>
> --
> David Lindsay
> anti spam in force use reply to address