Clustering groups according to equal standard deviation

[drfg2008]

I would like to cluster (order) groups according to equal standard deviations. So, instead of ordering (clustering) groups according to the mean as it is the case for example in the Scheffé-Post Hoc Test Procedure, I would like to do that according to the standard deviation.  There are about 1.200 groups, each group is represented in the data-set about 10-30 times.
As an example: I have 1.200 Football-Teams. Each Football Team scores 0-x scores in each game. The Football Teams play 10 – 30 matches. The question: which of the Football-Teams have the same standard deviations (building groups of teams with equal standard deviations).
Any idea very welcome.
Frank

[Art Kendall]

If you just want to order the teams see AGGREGATE.� use OUTFILE = * ADDVARIABLES and SD(teamsd) = SD(something).
use use OUTFILE = * ADDVARIABLES� or name a specific dataset or file.

then use the visual binning or RANK to create groups.

Clustering on a single variable (SD) occasionally pops up in the literature.� You might search the archives of CLASS-L and post a request there if you do not find something. [hidden email]

Art Kendall
Social Research Consultants

On 11/16/2011 4:51 AM, drfg2008 wrote:

I would like to cluster (order) groups according to equal standard
deviations. So, instead of ordering (clustering) groups according to the
mean as it is the case for example in the Scheffé-Post Hoc Test Procedure, I
would like to do that according to the standard deviation.  There are about
1.200 groups, each group is represented in the data-set about 10-30 times.
As an example: I have 1.200 Football-Teams. Each Football Team scores 0-x
scores in each game. The Football Teams play 10 – 30 matches. The question:
which of the Football-Teams have the same standard deviations (building
groups of teams with equal standard deviations).
Any idea very welcome.
Frank


-----
Dr. Frank Gaeth
FU-Berlin

--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Clustering-groups-according-to-equal-standard-deviation-tp4997226p4997226.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD

[David Marso]

AGGREGATE then RANK?
--

drfg2008 wrote

I would like to cluster (order) groups according to equal standard deviations. So, instead of ordering (clustering) groups according to the mean as it is the case for example in the Scheffé-Post Hoc Test Procedure, I would like to do that according to the standard deviation.  There are about 1.200 groups, each group is represented in the data-set about 10-30 times.
As an example: I have 1.200 Football-Teams. Each Football Team scores 0-x scores in each game. The Football Teams play 10 – 30 matches. The question: which of the Football-Teams have the same standard deviations (building groups of teams with equal standard deviations).
Any idea very welcome.
Frank

[drfg2008]

it's probably a contradiction in itself: to 'cluster' so many entities (teams) on a basis of significance tests (alpha error accumulation). Thought about something like a sort of Scheffe test just not for differences of the mean µ, but for sigma. I'm going to check CLASS-L, thanks.

Frank