1
$\begingroup$

The goal is to have groups emerge bottom-up from the data.

I have a number of measures (variables A, B, C, D), each with a value 1-10, for a large number of subjects.

Now, knowing that there will be different patterns of correlation between the variables for different subjects, I would like to know which subjects exhibit the same pattern.

In other words, what sort of analysis would reveal a) those differences in correlational patterns, and b) which subjects exhibit fall into which pattern?

Is such an analysis at all possible? And, if so, how would one go about it?

Improve this question
$\endgroup$
1
  • 1
    $\begingroup$ If you're looking to group your observations (i.e. subjects) on the basis of similarity then you may want to look at cluster analysis. $\endgroup$
    – Ian_Fin
    Commented Mar 22, 2017 at 18:42

1 Answer 1

Reset to default
-1
$\begingroup$

To create groups based upon correlations, you can create a matrix plot and look for various types of correlation within your data. Depending on the software you are using, you should be able to add regression lines and obtain a table of all of the Pearson's correlations for each pair.

If you are only looking for linear correlations, you may also want to look into a correlation matrix plot of your data.

Improve this answer
$\endgroup$
1
  • $\begingroup$ As I understand the question, @Thomas wants to group the subjects according to correlations. So, for example, maybe for half the subjects X positively correlates with Y, and for the other half it doesn't. What you're describing seems to aggregate across subjects, so you wouldn't necessarily see this, and seems closer to factor analysis, grouping of variables, rather than grouping of subjects. $\endgroup$
    – Ian_Fin
    Commented Mar 23, 2017 at 9:17

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.