Turn on suggestions

Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

- JMP User Community
- :
- Discussions
- :
- Discussions
- :
- Scalling data before cluster analysis

Topic Options

- Start Article
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Jul 6, 2018 6:00 AM
(783 views)

Hello everybody,

I have experimental data of patient cohort of 20 samples. For each patient I have a set of paramters. This paarameters are of different scalling, some of ELIZA dat measured with rang of 0-6000 pg/ml while other like T cell percentage range 0-100. I would like to run a cluster alysis for my chohort to sub group my sample due to different diseas stages for exmple. My question is that, do I need to re scalle the data before running such clustering with JMP. because as fa as I see that people are re scalleing their data (with Z score for example) before they cluster with python or R.. etc. If I need to rescalle the data, what would be better approach for that.

Thanks in advance for helpeing.

1 ACCEPTED SOLUTION

Accepted Solutions

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

I assume that you intend to use hierarchical clustering (unsupervised) to determine the number and identify of the clusters. Centering or scaling is not required for this method. Other multivariate methods benefit from centering and scaling.

Learn it once, use it forever!

3 REPLIES 3

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

I assume that you intend to use hierarchical clustering (unsupervised) to determine the number and identify of the clusters. Centering or scaling is not required for this method. Other multivariate methods benefit from centering and scaling.

Learn it once, use it forever!

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Re: Scalling data before cluster analysis

Dear Mark,

Thanks for the fast reply. Yes, me and my suprvisor consider hierarchical clustering without pre assumption for number of clusters. But what about if we consider to cluster using K mean (of courcewith pre assumption for cluster number)? Do we also need no scalling?

Regards

Ahmed

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Re: Scalling data before cluster analysis

K Means Clustering provides scaling and this option is enabled by default when you launch the platform.

Learn it once, use it forever!