7
$\begingroup$

In a psychological paper (Fujita et al., 2006), the authors perform a two sample t-test for independent means. They report the following values:

$M_1$ = 9.88, $M_2$ = 8.47,

t(66) = 2.25, p = 0.03, $p_{rep}$ = .94, r = .27

What does r mean in this context? Is it a correlation coefficient or is it an effect size? In the back of my head, I remember that r can be an effect size. I wish to know the effect size. What is the relation between the effect size and correlation? gives me some understanding, but how exactly can I interpret this r or convert it do Cohen's d?

For those who wonder what $p_{rep}$ is: wiki says it is a bad measure, nowadays its conception is proven to contain mathematical errors.

(The paper is behind a paywall)

Fujita, K., Henderson, M. D., Eng, J., Trope, Y., & Liberman, N. (2006). Spatial Distance and Mental Construal of Social Events. Psychological Science, 17(4), 278-282. https://doi.org/10.1111/j.1467-9280.2006.01698.x
Improve this question
$\endgroup$
7
  • $\begingroup$ Please provide a full citation for (Fujita et al., 2006). Then -- if the paper is not behind a paywall -- the answers won't have to make guesses about the "usual" and "without further information". $\endgroup$
    – dipetkov
    Commented Nov 23, 2023 at 12:34
  • 1
    $\begingroup$ @dipetkov I looked it up and the paper is behind a paywall, sorry. But actually the paper is not important though, because I included everything that is relevant in my question. It could be about any paper that reports $r$ along with a t-test. You have to take my word that there is nothing else that is useful in this paper to answer this question. $r$ is just dropped after a $t$ statistic and $p$ value, no explanations. $\endgroup$
    – uke
    Commented Nov 23, 2023 at 21:41
  • 1
    $\begingroup$ i added the reference. And what you are saying, we won't even compute the correlation, is interesting. Why not? The important thing, that it is a two independent samples scenario was clearly stated in my question though. $\endgroup$
    – uke
    Commented Nov 24, 2023 at 9:39
  • 1
    $\begingroup$ Thank you. Correlation is computed between two measurements taken from the same sample / person / object. Say we have two groups, men and women; we can't "align" those to compute a sample correlation coefficient; see wikipedia. Unless perhaps we have a sample of partnered couples; then we can pair those. $\endgroup$
    – dipetkov
    Commented Nov 24, 2023 at 10:10
  • 1
    $\begingroup$ It's true, your question is clear & the thing about proper citing of sources is something of a pet peeve of mine. What I noticed is that the two answers are incompatible but I admit that's not due to the content of your question. $\endgroup$
    – dipetkov
    Commented Nov 24, 2023 at 10:13

2 Answers 2

Reset to default
9
$\begingroup$

The reported $r$ is usually referring to an effect size. It is calculated from the t-statistic and the degrees of freedom

$$ r = \sqrt{\frac{t^2}{t^2 + df}} $$

where $t$ is the t-statistic and $df$ is the degrees of freedom ($N-2$) ​

​

Improve this answer
$\endgroup$
4
  • $\begingroup$ is this $r$ equivalent to calculating a certain correlation directly from the data? I wonder: is it possible to transform this formula to the formula of some correlation coefficient? $\endgroup$
    – uke
    Commented Nov 24, 2023 at 9:51
  • 1
    $\begingroup$ As already commented, a correlation is only defined if the data are paired. $\endgroup$
    – Nick Cox
    Commented Nov 24, 2023 at 10:06
  • 2
    $\begingroup$ Interesting, does this effect size have a name? Since usually effect sizes aim to provide a standardized measure of effect which is independent of sample size (in contrast to quantities such as the p-value), it seems a bit odd that $df$ is included in the formula. $\endgroup$
    – LuckyPal
    Commented Nov 24, 2023 at 10:56
  • $\begingroup$ @LuckyPal I think it is about removing the influence of sample size ($df$) from the test statistic ($t$ in this case), that's why it is in the denominator. $\endgroup$
    – uke
    Commented Feb 23 at 15:49
9
$\begingroup$

Without any other information, I would say $r$ is correlation. And that is an effect size.

Improve this answer
$\endgroup$

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.