Presenting Euclidean Distances

[Adam Troy]

Hi all,  

I have a table of Euclidean Distances for 100 cases (a 100 x 100 table) which I computed using PROXIMITIES. I am looking to present this data visually to identify any clusters of cases that represent groups of similar cases.  

Is there a preferred method of presenting a visual map of this data that can be implemented with SPSS 20?  

Thanks,

Adam Troy

[Zuluaga, Juan]

I may not be understanding the question - what is wrong with Multidimensional scaling plots (PROXSCAL) or with dendrograms?

"I have a table of Euclidean Distances for 100 cases (a 100 x 100 table) which I computed using PROXIMITIES. I am looking to present this data visually to identify any clusters of cases that represent groups of similar cases.

Is there a preferred method of presenting a visual map of this data that can be implemented with SPSS 20?"

=====================
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]

Nobody wants to stare at a 100x100 table .  Even if such a person existed the table would be uninterpretable.   Natural ways of presenting such would be through a dendogram (see CLUSTER command which reads the previously generated distance matrix created in PROXIMITIES ).  I believe you can also get plots through MDS (ALSCAL) but I have not used that in some time.  Depending upon the nature of the data, correspondence analysis might provide additional insights.
---

Adam B. Troy-3 wrote

Hi all,

I have a table of Euclidean Distances for 100 cases (a 100 x 100 table)
which I computed using PROXIMITIES. I am looking to present this data
visually to identify any clusters of cases that represent groups of similar
cases.

Is there a preferred method of presenting a visual map of this data that
can be implemented with SPSS 20?

Thanks,

Adam Troy

[Art Kendall]

The answer to that question depends to some degree on the nature of your data and who your audience(s) are.

What constitutes a case? I.e., what are the distances between?

What are the dimensions (variables usually) that went into the distances?  What were they? How were they chosen? How many are they?

Why did you choose Euclidean distance?  Is there a reason to assume that the dimensions are pretty much uncorrelated?

Do you expect to find a tree structure? Or are you looking to construct a single nominal level variable whose levels are the clusters?


On 7/16/2012 11:08 AM, Adam Troy wrote:

Hi all,  

I have a table of Euclidean Distances for 100 cases (a 100 x 100 table) which I computed using PROXIMITIES. I am looking to present this data visually to identify any clusters of cases that represent groups of similar cases.  

Is there a preferred method of presenting a visual map of this data that can be implemented with SPSS 20?  

Thanks,

Adam Troy

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

[Adam Troy]

Thanks everyone. On further investigation it looks like multidimensional scaling works quite nicely for this purpose.

Adam

On Wed, Jul 18, 2012 at 7:54 AM, Art Kendall <[hidden email]> wrote:

The answer to that question depends to some degree on the nature of your data and who your audience(s) are.

What constitutes a case? I.e., what are the distances between?

What are the dimensions (variables usually) that went into the distances?  What were they? How were they chosen? How many are they?

Why did you choose Euclidean distance?  Is there a reason to assume that the dimensions are pretty much uncorrelated?

Do you expect to find a tree structure? Or are you looking to construct a single nominal level variable whose levels are the clusters?

Art Kendall
Social Research Consultants

On 7/16/2012 11:08 AM, Adam Troy wrote:

Hi all,  

I have a table of Euclidean Distances for 100 cases (a 100 x 100 table) which I computed using PROXIMITIES. I am looking to present this data visually to identify any clusters of cases that represent groups of similar cases.  

Is there a preferred method of presenting a visual map of this data that can be implemented with SPSS 20?  

Thanks,

Adam Troy