1
$\begingroup$

Hope someone can help me here:

  • So there are variables A, B, C, and D and groups 1 and 2.

  • Groups significantly differ in variables A and B, but not in C and D.

  • From previous research, correlations are generally interesting for (1) A and B, (2) and C and B (3) and C and D

(1) - When conducting partial correlation between A and B, controlling for group, no significant result is to be obtained/ or one with p= 0.052. - When splitting the file by means of group and conducting bivariate correlation, only second group has significant correlation of A and B (meaningful interpretation possible) --> is the split file proceedure okay here? And is an interpretation possible then?

(2) - When conducting partial correlation of C and B in overall sample, controlling for group, there is significant result - When splitting the file by means of group and conducting bivariate correlation, only second group has significant correlation of C and B (meaningful interpretation possible) --> is the split file proceedure okay here?

(3) - When conducted partial correlation of C and D in overall sample, controlling for group, there is significant result - When splitting the file by means of group and conducting bivariate correlation, only first group has significant correlation of C and D (meaningful interpretation possible) --> is the split file proceedure okay here?

I hope you have the rough idea of what I am trying to do and what the results are. The basic question is, can I just split a data set by factor group and look for correlations between variables that already differ between the groups?

I know that partial correlation is meant to control for any associations with the third variable (i.e. group) and is the proceedure of choice. But is splitting the data set absolutely not allowed here?

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.