topic How to show only significant correlation (>= .6) in a large correlation matrix in Discussions

How to show only significant correlation (>= .6) in a large correlation matrix

helalmob — Sun, 11 Jun 2023 11:22:11 GMT

I have 33 Likert scale variables that measure different types of disabilities from facial expression to speech, etc. The total score across all 33 scores gives a good index of an overall disability level. However, I found out that some items are highly correlated and perhaps redundant. I ran a correlation matrix for the 33 items in JMP Pro 16 that produced a huge table with the majority of correlations >= .5 and some > .5. My question is how to only show > .5, or >= .6 correlation in the matrix? The scatterplot is more visual but need to select only the ones in the red circle and exclude the rest

Thank you,

Re: How to show only significant correlation (>= .6) in a large correlation matrix

Mark_Bailey — Thu, 24 Feb 2022 19:46:21 GMT

Click the red triangle at the top of Multivariate and select Pairwise Correlations. This report might be a better layout for you.

You can right click in this table and select Sort by Column. I would use p-value so that you do not end up with strong correlations at the top and at the bottom because of negative or positive correlation. You could also right click and select Make Into Data Table, and then create a new column with Abs( corr ) for sorting the whole table.

There are other solutions, but hopefully you get the idea.

Re: How to show only significant correlation (>= .6) in a large correlation matrix

helalmob — Thu, 24 Feb 2022 20:24:41 GMT

Thank you Mark. It worked perfectly and that's exactly what I needed.

Re: How to show only significant correlation (>= .6) in a large correlation matrix

txnelson — Thu, 24 Feb 2022 21:44:48 GMT

I think there is an alternative approach that you may find interesting. I find the Response Screening platform provides a great interactive platform for investigation of correlation results. It creates a data table of all of the results, with an additional window of plots and charts for the interactive viewing of the comparisons.

Try it out, it may give you a new way to look at the data