topic Re: JMP 17.2 > Variable Clustering > Algorithm and Output Display? in Discussions

JMP 17.2 > Variable Clustering > Algorithm and Output Display?

Thierry_S — Tue, 19 Mar 2024 17:29:16 GMT

Hi JMP Community,

Windows 10 Pro, JMP 17.2

I am looking for an approach to display graphically the output of the Cluster Variable platform.

I could recluster the variables using the general clustering platform and generate the associated dendrogram. However, I don't know which specific algorithm the Cluster Variables platform uses (Ward's method does not match the Cluster Variables output).
I could potentially rebuild a dendrogram from the output of the Cluster Variable platform, but I must admit that I am unsure about the best approach to achieve this goal.

All suggestions are welcome.

Thank you.

Best,

Re: JMP 17.2 > Variable Clustering > Algorithm and Output Display?

Jed_Campbell — Tue, 19 Mar 2024 19:18:34 GMT

I'm not really certain what type of visualization you're looking for, but I like the Parallel Plots as a way to see the clusters. These are available in most (all?) of the clustering platforms, so you don't have to do it by hand. I've attached a data table with an example.

Re: JMP 17.2 > Variable Clustering > Algorithm and Output Display?

Thierry_S — Tue, 19 Mar 2024 20:31:24 GMT

Hi Jed,

Thank you for your answer. I am unsure how the Parallel Plot platform could be used on the Cluster Variable platform output, but I will explore it.

What I would like to achieve is to display the different clusters and their components as a dendrogram.

Thanks.

Best regards,

Re: JMP 17.2 > Variable Clustering > Algorithm and Output Display?

Jed_Campbell — Tue, 19 Mar 2024 20:52:10 GMT

Ah, perhaps Red Triangle...Two Way Clustering would work for you. It shows the components (as colors) next to the dendrogram.

Re: JMP 17.2 > Variable Clustering > Algorithm and Output Display?

Thierry_S — Wed, 20 Mar 2024 05:16:10 GMT

Hi Jed,

Sorry for not explaining myself clearly. I am not talking about the Hierarchical or K Mean clustering platforms. I specifically referring to the CLUSTER VARIABLE platform. I am well aware of the multiple options in the Hierarchical or K Mean clustering platforms.

My questions are:

What is the algorithm used by the CLUSTER VARIABLE platform
Alternatively, how can I visualize (dendrogram) the relationship between clusters and members produced by the CLUSTER VARIABLE platform?

Thank you.

Best,

Re: JMP 17.2 > Variable Clustering > Algorithm and Output Display?

Jed_Campbell — Wed, 20 Mar 2024 14:43:47 GMT

Hi Thierry,

Sorry, I see that you were clear from the start about Cluster Variables platform, but I just wasn't reading properly. I found 2 ways to make an almost dendrogram, perhaps others in the community can improve these or provide better ideas. I started by making the Cluster Members table into a data table. The first method was a Diagram:

To force this into just one diagram instead of a separate one for each cluster, I added a '0' cluster to the bottom of the data table:

I also made something in Graph Builder:

To get this, I had to add a 'Main' column to the data table, and I had to create a manual Value Ordering sort order for the Members column.

I attached a data table, and I'm hoping others might have better ideas (perhaps graphing guru @scwise).

Thanks for helping me see what you're looking for.

Jed

Re: JMP 17.2 > Variable Clustering > Algorithm and Output Display?

Mark_Bailey — Wed, 20 Mar 2024 17:40:38 GMT

Have you read Statistical Details in the documentation for the Variable Clustering?

Re: JMP 17.2 > Variable Clustering > Algorithm and Output Display?

Thierry_S — Thu, 21 Mar 2024 04:12:08 GMT

Yes, but I may have missed the specific information I seek.

Specifically, I could not find the clustering algorithm used by the CLUSTER VARIABLE platform.