use developed clusters to classify future patients

[Matthew Pirritano]

SPSS’ers,

 

I have used cluster analysis to classify patients into 6 clusters based on psychological constructs. How could a clinician assess people on these constructs and then assign them to a cluster with some degree of reliability?

 

Do I need to create a regression type model using Discriminant Function Analysis? There must be some standard way of doing this but I’m having trouble finding any literature, and there was only one   correspondence in the list archives that hinted at a method for creating a “post-classification” model, which sounds like what I want to do.

 

Any help is much appreciated!

 

Thanks

Matt

[Rich Ulrich]

I believe you have just discovered that cluster analysis is not especially
fruitful.  It does not take you very far in the direction you want to go.
You are into "data-mining"  and clustering is not much of a start.

The results that I've heard of, assigning to groups with a specified degree
of reliability, are in anthropology -- I'm thinking of data where the original
data base includes many bones whose identity (species, age, sex, etc.)
is known, so those can be problems that are easier than yours. 

I've never looked for a web source for that sort of problem, though.

The website that I know of that discusses related methods is one that
has a special focus on ecology.  The is a site on "ordination" at
Oklahoma State University. 
    http://ordination.okstate.edu/overview.htm

I've used the site, for instance, to learn more about Correspondence
Analysis.  It may help your perspective on techniques, once you have
particular programs in mind.  Right off, I don't see anything there
about ordinary clustering.

--
Rich Ulrich

---

Date: Sat, 12 Nov 2011 17:55:06 -0800
From: [hidden email]
Subject: use developed clusters to classify future patients
To: [hidden email]

SPSS’ers,

 

I have used cluster analysis to classify patients into 6 clusters based on psychological constructs. How could a clinician assess people on these constructs and then assign them to a cluster with some degree of reliability?

 

Do I need to create a regression type model using Discriminant Function Analysis? There must be some standard way of doing this but I’m having trouble finding any literature, and there was only one   correspondence in the list archives that hinted at a method for creating a “post-classification” model, which sounds like what I want to do.

 

Any help is much appreciated!

 

Thanks

Matt

[news]

Matt,

What about using the clusters as dependent variable in a discriminant
analysis ? I have not used DA for a long time but I am sure that there
is an option to classify unclassified cases.

HTH

F. Thomas

On 13/11/2011 02:55, Matthew Pirritano wrote:

> SPSS’ers,
>
> I have used cluster analysis to classify patients into 6 clusters based
> on psychological constructs. How could a clinician assess people on
> these constructs and then assign them to a cluster with some degree of
> reliability?
>
> Do I need to create a regression type model using Discriminant Function
> Analysis? There must be some standard way of doing this but I’m having
> trouble finding any literature, and there was only one correspondence in
> the list archives that hinted at a method for creating a
> “post-classification” model, which sounds like what I want to do.
>
> Any help is much appreciated!
>
> Thanks
>
> Matt
>

=====================
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

[Jon K Peck]

Statistics includes a scoring function named APPLYMODEL that can take a TWOSTEP CLUSTER result and use it to score new cases.  APPLYMODEL is used with COMPUTE.  Originally APPLYMODEL was limited to the Server version, but recent version(s) include it in the Client.  There is now a Scoring Wizard on the Utilities menu that helps you generate the appropriate syntax.

HTH

Jon Peck (no "h") aka Kim
Senior Software Engineer, IBM
[hidden email]
new phone: [phone]




From:        ftr <[hidden email]>
To:        [hidden email]
Date:        11/21/2011 12:32 PM
Subject:        Re: [SPSSX-L] use developed clusters to classify future patients
Sent by:        "SPSSX(r) Discussion" <[hidden email]>

---

Matt,

What about using the clusters as dependent variable in a discriminant
analysis ? I have not used DA for a long time but I am sure that there
is an option to classify unclassified cases.

HTH

F. Thomas

On 13/11/2011 02:55, Matthew Pirritano wrote:
> SPSS’ers,
>
> I have used cluster analysis to classify patients into 6 clusters based
> on psychological constructs. How could a clinician assess people on
> these constructs and then assign them to a cluster with some degree of
> reliability?
>
> Do I need to create a regression type model using Discriminant Function
> Analysis? There must be some standard way of doing this but I’m having
> trouble finding any literature, and there was only one correspondence in
> the list archives that hinted at a method for creating a
> “post-classification” model, which sounds like what I want to do.
>
> Any help is much appreciated!
>
> Thanks
>
> Matt
>

=====================
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