2
$\begingroup$

I am performing an independent t-test, in which the independent variable is the "group" which has two values A and B representing an approach the participants used, and the dependent variable is a metric for accuracy "Recall" which has numeric values ranging from 0 to 100. Total number of participants is 18 (9 in each group) randomly assigned to the groups.

What is the formula I should use to use in calculating the effect size in terms of the Point-Biserial Correlation (r), knowing that the two independent variables are non homogeneous according to Levene's test?

$\endgroup$
3
  • $\begingroup$ What variable(s) are ordinal - DV or IV? You say two independent variables what do you mean? "Independent T-test" is done with one grouping independent variable with 2 groups. $\endgroup$
    – ttnphns
    Commented Feb 3, 2016 at 20:44
  • $\begingroup$ Hello ttnphns, the IV is the "group" which has two values A and B representing an approach they used, the DV is metric for accuracy "Recall" which has numeric values ranging from 0 to 100. $\endgroup$
    – H.A.
    Commented Feb 4, 2016 at 11:09
  • $\begingroup$ How can you possibly calculate a correlation coefficient on independent data? Correlation measures are for pairs/dependent groups. $\endgroup$
    – Alexis
    Commented Jun 30, 2018 at 0:14

1 Answer 1

2
$\begingroup$

You realise that the point-biserial is numerically equivalent to the Pearson, right? You can look at the wikipedia page or use an online calculator.

It sounds odd to me to be calculating a correlation between different people in the two groups. Correlation always works on paired observations. Do you not just want an effect size using something like Cohen's d?

$\endgroup$
1
  • $\begingroup$ Hello stan, I have already calculated Cohen's d, but another part of the study was about the "precision" variable which was not normal and I had to use the Mann-Whitney test and I had to report effect size as r=z/sqrt(n). So, I thought I should have similar types of effect size reported for both recall and precision. $\endgroup$
    – H.A.
    Commented Feb 5, 2016 at 15:32

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.