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
- :
- Consider only certain combinations when doing Multivariate Analysis

Topic Options

- 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

Aug 12, 2013 9:39 AM
(1157 views)

Hi all

I have data for four different variables: A, B, C and D. Now I want to do a Multivariate Analysis using the following steps in JMP: Analyze --> Multivariate Methods --> Multivariate. Then I select my four variables and add it as Y, Columns. After that, I choose Spearman's p in the nonparametric correlations menu. Now I see the correlations between all 4 variables.

Now my problem is that I would like to limit the number of correlations that I see. I want only to see the correlations between A and C, A and D and also between B and C, respectively B and D. I can't find any option for this. Is there any way how to achieve that?

Regards

SM

1 REPLY

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

Hi SM,

There is no option in the Multivariate platform to perform a subset of the correlations as you describe.

However, you can use the Fit Y by X platform to get the pairs that you're looking for. However, since you're looking for Spearman's ρ, you'll need to use the ranks of the actual values in Fit Y by X platform.

Here's an example using the Decathlon data from JMP's Sample Data Library (Help->Sample Data, then look in the Multivariate section).

Here's the first part of the Decathlon data.

If I want the non-parametric correlations between Score and Long Jump, Score and Shot Put, 100m and Long Jump and 100m and Shot Put then first I need the rank of each row for each of those columns. This is easy to do using the Distribution platform.

Launch the platform with these four columns in the Y role.

Then, hold down the Ctrl key (Command key on the Mac) as you choose Save->Ranks Averaged from the red triangle menu at the top of any of the histograms.

It's important to hold the Ctrl key to "broadcast" that menu choice to all of the column in the analysis.

You'll get four new columns in your data table.

Use these new columns in Fit Y by X.

You'll get a report window with each Y against each X (four in total for this case).

Again, hold down the Ctrl key as you choose Density Ellipse->0.95 in the red triangle menu above any scatterplot.

You'll get a Correlation report below each scatterplot. The Correlation and Signif. Prob columns in this report are Spearman's ρ and it's p-value for that pair.

If you would like a data table of all of these values for all the pairs in the report, you can right click on the report table and choose "Make Combined Data Table" from the menu.

You'll get a data table like this.

I hope that helps. Let us know if you have any further questions or difficulties.

-Jeff

-Jeff