How do I create clustering tree/view with a distance matrix?

Question

I am trying to visualize the organization of some data with JMP9.

I have N samples and I have calculated a distance matrix of N*N with my own formula. For example, I have shown some made up 4X4 comparison matrix here.

10012     10013     10014     10015

10012    0    100    100    200

10013    2    0    300    100

10014    300    100    0    33

10015    30    200    38    0

I opened this matrix in JMP. Then I tried to use the Analyze->Multivariant -> Cluster. Then I clicked the "Data is Distance Matrix" and clicked OK button. Then nothing happened. Exactly how do I do it? Anyone with any experience?

Thanks,

Heng

hengdai · Answer

I tried and it worked.  The reason that I failed early was that I forgot to label ColName column as "char" since it was identified as numeric by default.

Thanks.

unknown · Answer

Hi - I've just tried it myself: it works for me in both JMP 9 and JMP 10.  You'll need to load your data into a data table as five columns called (for example) "ColName", "10012", "10013", "10014" and "10015" respectively, with each column consisting of four rows, and ColName containing the four numbers 10012, 10013, 10014 and 10015 (because the column containing the identifiers needs to correspond to the names of the columns containing the actual data).  Then just call up Analyze | Multivariate Methods | Cluster, click on "Data is distance matrix" as you've already described, and identify column "ColName" as the Label column in the resulting dialog screen.  You should get a warning to the effect that your distance matrix isn't symmetric, but it will actually be analyzed.

I've tried entering the column identifiers as both numeric and alphabetic: it doesn't seem to make any difference.

Does that help?