I want to explore if there is a gender difference in a correlation between two ordinal items. I have calculated spearman rho within gender for the items, but I now need to test if the correlation coefficient is different between men and woman. Is there a way I can do this in SPSS?
Hello Terans Gunawardhana. Your suggestion to compute a point-biserial correlation makes me wonder if you understood Veronica to be computing the correlation between sex and a continuous variable. If so, I understood things differently. I believe she is computing the Spearman correlation between X and Y for males and females separately, and she wants to compare those two correlations. HTH.
SPSS has an option split file. Once you have split the file for gender then it will give separate results for male and female, but finding whether they differ will still be a subjective matter. Using gender as a dummy to run regression will be more appropriate.
Hello Veronica L Meredith. The Spearman correlation is the same thing as Pearson's r computed on ranks. So assuming you have reasonably good sample sizes, I believe you can use the usual method for comparing two independent Pearson correlations. As you are using SPSS, see syntax file #3 here:
- https://sites.google.com/a/lakeheadu.ca/bweaver/Home/statistics/spss/my-spss-page/weaver_wuensch
See also this older discussion of the same question, and note the reference to Jerrold (2005) in Encyclopedia of Biostatistics.
- https://www.researchgate.net/post/how_do_you_compare_spearmans_correlations_correlations_corresponding_to_ordinal_data_of_2_independent_groups
HTH.
Dear Veronica L Meredith ,
See whether you can use Point-Biserial Correlation.
Good Luck!
Terans
Hello Terans Gunawardhana. Your suggestion to compute a point-biserial correlation makes me wonder if you understood Veronica to be computing the correlation between sex and a continuous variable. If so, I understood things differently. I believe she is computing the Spearman correlation between X and Y for males and females separately, and she wants to compare those two correlations. HTH.
If you are interested in the differential influence of gender, how about an ordinal regression and you include gender and the genderXPredictor interaction into the model. Otherwise go with Bruce's suggestion.
I agree with Bruce Weaver I had misunderstood the question raised by Veronica L Meredith
Correlation can be calculated for the two groups separately. Significance will come with the test. But no way to test the significance of the difference of correlation. The regression will give some idea but it is not exacly the correlation difference.
@Dileepa why do you think a difference of the cortelation coefficients is not possible? Please provide any sources. Bruce's idea and sources to treat the Spearman correlation like a Pearson and use Fisher's z-transformation seems very reasonable. What speaks against it per se (despite possible power issues)?
I mean no direct method. There are some complicated methods through GLM.
see the link
https://www.ibm.com/support/pages/testing-group-difference-correlation-glm-approach
@Dileepa how about this link to do it in SPSS via Syntax
https://www.ibm.com/support/pages/differences-between-correlations
Or directly use an online calculator like :
https://www.psychometrica.de/correlation.html
PS in the first Link is also a solution for Spearman rank correlations.
@ Rainer, thank you for the online calculator. I never used it. I used GLM method always.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS