1
$\begingroup$

[This may have been asked before, but I'm on the android app and the search function seems unable to find it]

I'm getting confused with p-values and the null hypotheses. I'm testing if two values are related and I've done some reading and decided to use spearmans rank correlation. An example of a value I get from one of my tests is as follows:

Spearmans: 0.07 P-val: 0

The p-val isnt always exactly 0, sometimes is a tiny number (eg. 0.0007). Now I understand that a spearmans of 0.07 is a very weak correlation, but does a p-value of near zero imply that the correlation is very statistically significant?

My understand of the above would mean that we can be very confident that the weak correlation is present; ie. Changing value 1 will result in a change in value 2, but the amount value 2 changes is hard to predict.

$\endgroup$

1 Answer 1

1
$\begingroup$

You've nailed it.

Effect size and p-value are related, but sample size also some into play when determining the p-value.

It is totally reasonable to think that, with a very large sample size, you can say, with extreme confidence, that your quantities have a nonzero Spearman correlation. This is because of the tiny p-value giving strong evidence against a null hypothesis of zero Spearman correlation.

At the same time, $ 0.07$ is quite weak correlation, likely not of practical significance to your problem. The reason you are getting a small p-value is because of a large sample size saying that $0.07 \ne 0$.

$\endgroup$

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.