multivariate correlation

[#1]

Hi, is there a way to filter (corr>x, or like top 10 corr etc) or to at list to sort  the correlation when i am analyzing  the multivariate correlation results? 

thank you!

Hillel

[#2 · Accepted solution]

You may want to try thr  Response Screening platform. 

     Analyze=>Screening=>Response Screening

It is designed to sort and select the results based upon the te significance level.

[#3 · Accepted solution]

Hi @hillel, welcome to the forum.

@txnelson gave you by far the most powerful approach.
Perhaps a more simple approach can be the following:

produce the correlations using the multivariate platform. There, ask for the pairwise correlations. Right click on the pairwise correlation table and select >> Sort table >> By Pvalues >> ascending.

using a script its should look like this:

dt = Open( "$SAMPLE_DATA\semiconductor capability.jmp" );


Multivariate(
	Y( :NPN1, :PNP1, :PNP2, :NPN2, :PNP3, :IVP1, :PNP4 ),
	Estimation Method( "Row-wise" ),
	Scatterplot Matrix( 1 ),
	Pairwise Correlations( 1 ),
	SendToReport(
		Dispatch( {}, "Correlations", OutlineBox, {Close( 1 )} ),
		Dispatch(
			{"Pairwise Correlations"},
			"",
			TableBox,
			{Sort By Column( 7, 1 )}
		)
	)
);

You can also make a data table from the pairwise correlations and take things further.

Let us know if it helps,

Ron

[#4]

This feature works with any table presented in a JMP platform: First, select Pairwise Correlations from the red triangle menu next to Multivariate. Then right-click in the table and choose Sort By Column. Select Signif Prob and Ascending and click OK. Now the list is sorted from most significant to least significant. If you want to sort by correlation, then right-click and select Make Into Data Table. Right-click on the Correlation data column and select New Column Formula. Choose Transform and Absolute. Now click Tables > Sort to finish.

[#5]

The other posters have given you multiple ways to get what you want. My only advice is to look at the data and don’t just rely on the quantitative r and p-values. These statistics can be influenced by singular data points and “ outliers”.