topic Filtering through Scatterplot Matrix to find significant interactions in Discussions

Filtering through Scatterplot Matrix to find significant interactions

ShreyasP — Sun, 11 Jun 2023 11:24:21 GMT

I have a large data set and I am interested in visualizing unique interactions. I made a scatterplot matrix displayed with fit lines and 95% confidence intervals. Is there a way to filter the matrix to select for or display interactions by significance or p-value?

Re: Filtering through Scatterplot Matrix to find significant interactions

txnelson — Wed, 18 May 2022 20:38:46 GMT

The Response Screening Platform

Analyze=>Screening=>Response Screening

The platform produces a data table with the pvalues, which will allow you to sort them, and using the built in scripts, you will be able to display their plots.

Re: Filtering through Scatterplot Matrix to find significant interactions

Phil_Kay — Thu, 19 May 2022 09:32:29 GMT

Hi,

@txnelson 's suggestion is a good one. The Response Screening platform is very powerful.

However, I am not sure if that is what you need. I don't think you mean "interactions" - interaction effects are where one predictor variable influences the effect of another predictor variable on the response variable.

I think you really just mean that you want to find significant correlations out of all pairwise correlations of your variables.

You can do this with the Multivariate platform (Analyze > Multivariate Methods > Multivariate). Add your variables in the Y, Columns role as you did for the scatterplot matrix and click OK.

You can bring up a list of all pairwise correlations, including p-values, from the red triangle menu for Multivariate:

By right-clicking on the column headers you have the option to sort by the p-value column:

If you right-click on this table you can "Make into Data Table", which will make this into a new jmp table, if that is useful.

I hope that helps.

Phil