Dear all,
I am currently working on a study, in which I want to use SEM in order to examine the mediating effect of a variable in the relationship between other two variables. Before I use SEM, I want to perform a correlation between the composite variables in order to explore their relationship. However, many of my variables are not normally distributed, so I just want to know if I should perform a Pearson or spearman correlation before I start to use SEM? Many thanks in advance for your help,
Filipa Calado
Technically, Pearson correlation assumes that the data are bivariate normal, not that each variable is normal.
Spearman may be able to better pick up correlations that are relatively monotonic, but not linear. For example, for the attached plot, Pearson (p = 0.3, r = 0.37) and Spearman (p < 0.001, rho = 1.0).
My advice is to plot out all the correlations, eg.
http://rcompanion.org/handbook/images/image136.png
For the data shown there, the Pearson and Spearman correlations all come out about the same.
Dear Salvatore,
Thank you very much for your response and for your help. However, could you please tell me how I check in SPSS that the data are bivariate normal? Many thanks
This link is probably helpful.
https://statistics.laerd.com/spss-tutorials/pearsons-product-moment-correlation-using-spss-statistics.php
- Badges
- Science topic
- Similar topics
- Physical Sciences
- Materials Science
- Materials
- Inorganic Materials
- Ceramics