0
$\begingroup$

It seems to me that the various options for visualizing the correlation matrix in R are quite unintuitive for laymen. They focus on the graphical representation of the correlation matrix as different colors or shapes of the correlation coefficient - and I am thinking of representing them in terms of distance from each other. We visualize the distances between points, for example, using factor analysis or other analyses. How to create a graph/plot from a ready-made symmetrical correlation matrix in which the greater the distance, the smaller the correlation? What data transformation to use?

For start, because I want to show the "similarity" between the variables, I have such an assumption that the correlation at level 1 is the highest correlation and the distance from the points is zero, and the correlation -1 is the lowest correlation, i.e. the maximum distance from the points. But I'll also settle for a graph where negative correlations will be treated the same as positive correlations. I am asking for an example in R, because only in it I can create visualizations, or just the name of the technique that I could search for. Thanks in advance for all hints.

Improve this question
$\endgroup$
2
  • 3
    $\begingroup$ Biplots from principal component analysis (e.g. stats.stackexchange.com/questions/137240/…) are often used for this purpose. Tastes vary, but to me a scatter plot matrix is more informative than anything like a biplot -- which rests entirely on correlation being an adequate summary of each bivariate relationship. $\endgroup$
    – Nick Cox
    Commented Sep 4, 2023 at 19:59
  • 1
    $\begingroup$ It sounds like you are asking for an (approximate) isometric embedding of points on a high-dimensional sphere into the plane. Most of the time that won't work well; but if it might, various MDS methods should be able to find one. $\endgroup$
    – whuber ♦
    Commented Sep 5, 2023 at 14:23

0

Reset to default

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.