4
$\begingroup$

This question is inspired by this comment.

From this question, we know that zero correlation doesn't always mean two variables are related, so my question now is, how to ensure that there is absolutely no relationship whatsoever between two variables?

Improve this question
$\endgroup$
9
  • $\begingroup$ Is the sample size large? $\endgroup$
    – Michael M
    Commented Sep 28, 2015 at 14:25
  • $\begingroup$ @MichaelM, assume that the sample size is already sufficiently large $\endgroup$
    – Graviton
    Commented Sep 28, 2015 at 14:58
  • 2
    $\begingroup$ I don't think there is a way to prove independence. What you can do is examine as many possible dependences as you can and reject them. (Although be aware that 95% confidence level allow for 5% false positives.) $\endgroup$
    – Richard Hardy
    Commented Sep 28, 2015 at 20:28
  • 1
    $\begingroup$ The correctness of the answer will depend on what you mean by "variable" and "relationship." For instance, are you referring to data or to mathematical constructs? If the latter, then by "no relationship" do you mean independent or something else? $\endgroup$
    – whuber ♦
    Commented Sep 28, 2015 at 20:33
  • 2
    $\begingroup$ If you have data, there is no way to "ensure ... absolutely no relationship." There will always be "relationships" among the data; and you cannot ensure anything about a model from any finite amount of data. $\endgroup$
    – whuber ♦
    Commented Oct 6, 2015 at 3:21

0

Reset to default

Browse other questions tagged or ask your own question.