Statistically correct to compare two Spearman's ρ-Vaules directly?

G

gentix

Guest
Hello statistic-community :wave:

My sample size is approx. 1.500. I have different variables (A, B, C, D) which are all ordinal scaled. I have another variable (Z) which is ordinal scaled, too.

Is it - in a statistical sense - correct, to calculate the Spearman's ρ with SPSS for each relationship - (1) A & Z, (2) B & Z, (3) C & Z and so on - and compare these values subsequently, to get information about: Which relationship ist smaller, which ist stronger?

Maybe using the Fisher Z-transformation afterwards to look, whether the differences are significant?

Possible result: None of the relationships differs significantly from another one but (1) is slightly stronger than (2) but not as strong as (3)? Is that a conclusion, which makes sense? Or are there any statistical errors that block this conclusion?

greetings
gentix

Last edited by a moderator:

CB

Super Moderator
Hi there!

If I'm interpreting your question correctly, yes you can say that one relationship is stronger than the other, but that the difference is not statistically significant.

spunky

Doesn't actually exist
Maybe using the Fisher Z-transformation afterwards to look, whether the differences are significant?
This seems to still be an open problem as evidenced int his discussion. so just using Fisher's r-to-z transformation for the case of Spearman's rho is not quite correct. Notice that there are some guidelines on how to calculate and assess the statistical significance of this difference as long as the Spearman correlations are less than 0.6

Karabiner

TS Contributor
Moreover, these are correlated correlation coefficients from one and the same sample, which have one variable in common. In addition to the r-to-z transformation, one'd have to take into account the covariance of the estimates.

G

gentix

Guest
Thank you very much for your support. That's totally ok. I will go without Fisher-Z-transformation and compare the rho's without further information about statistical significance.

Greetings
gentix