0
$\begingroup$

I'm having a bit of difficulty grasping the concept of 'independence' for the purpose of testing whether two correlations significantly differ from one another (Zou, 2007).

Consider the following made-up example from a sample of adolescents randomly selected from a school:

  1. Student BMI (kg/m2) is correlated with a non-expert rating of diet quality (i.e., each individual has their diet rated by non-experts).

  2. Student BMI (kg/m2) is correlated with an expert rating of diet quality (i.e., each individual has their diet rated by experts).

In this instance, is it appropriate to assume that each correlation is independent of one another and come from separate groups (although BMI from the same individuals features in each correlation)? The purpose of this would be determine if each correlation significantly differs from one another using the Zou method.

$\endgroup$
0
$\begingroup$

I don't think these two are independent. The BMIs are (of course) the same in the two correlations and the ratings of diet are clearly related to each other. I mean a diet of fried chicken, Doritos and Coke is likely to get low ratings from just about anyone, expert or not.

For independence, it would have to be different people being rated.

$\endgroup$
1
  • $\begingroup$ Thank you for clarifying. It appears the two correlations are best characterised as overlapping dependent correlations. I believe this means I would need to calculate the correlation between non-expert and expert ratings for inclusion of a difference test using the Zou (2007) method. $\endgroup$ – pomodoro Feb 26 '18 at 13:42

Your Answer

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

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