help
help > RE: z score transform back to r value
Jun 30, 2015 02:06 AM | Alfonso Nieto-Castanon - Boston University
RE: z score transform back to r value
Hi Ivan,
Yes, that is the correct transformation (assuming that your second-level model is a one-sample t-test computing the average connectivity across all subjects), and r and Z are very similar for r values below .5 and only start to diverge more clearly for higher r values. While effect sizes vary considerably depending on the acquisition and preprocessing parameters and average correlation of .08 is not extremely low. Significance across subjects (in a one-sample t-test) is determined by comparing this average effect to the size of the between-subjects variability, so a significant effect means that this effect size is relatively large compared to the observed amount of between-subject variability in subject-specific connectivity values.
Hope this helps
Alfonso
Originally posted by Yang Yang:
Yes, that is the correct transformation (assuming that your second-level model is a one-sample t-test computing the average connectivity across all subjects), and r and Z are very similar for r values below .5 and only start to diverge more clearly for higher r values. While effect sizes vary considerably depending on the acquisition and preprocessing parameters and average correlation of .08 is not extremely low. Significance across subjects (in a one-sample t-test) is determined by comparing this average effect to the size of the between-subjects variability, so a significant effect means that this effect size is relatively large compared to the observed amount of between-subject variability in subject-specific connectivity values.
Hope this helps
Alfonso
Originally posted by Yang Yang:
Hi, Alfonso,
I hope to report r value in my paper. But when I use this r=tanh(Z), why the r values are almost the same as z values. If I transform the z score into r value, the result is 0.08 based on the transformation of r=tanh(z). Additionally, why a very low average z value (0.08) of a group of 15 participants could be significant (p<0.02) when I do one sample t test in post-hoc analysis. But, the relation coefficient seems to be so low. Is it a correct way to transfrom?
Thanks!
Ivan
I hope to report r value in my paper. But when I use this r=tanh(Z), why the r values are almost the same as z values. If I transform the z score into r value, the result is 0.08 based on the transformation of r=tanh(z). Additionally, why a very low average z value (0.08) of a group of 15 participants could be significant (p<0.02) when I do one sample t test in post-hoc analysis. But, the relation coefficient seems to be so low. Is it a correct way to transfrom?
Thanks!
Ivan
Threaded View
| Title | Author | Date |
|---|---|---|
| Yang Yang | Jun 24, 2015 | |
| Alfonso Nieto-Castanon | Jun 30, 2015 | |