The 'r' value obtained from partial correlation between x and y controlling for z is higher than that obtained from the bivariate correlation of x,y.
In another case, adjusted odds ratios from a binary logistic regression were higher compared to crude odds ratios.
This is called a suppressor effect. See e.g.:
Friedman, L., & Wall, M. (2005). Graphical Views of Suppression and Multicollinearity in Multiple Linear Regression. The American Statistician, 59(2), 127-136
It can. Whatever metric it is, in using Z -score and student-T score will give true location of performance on graph and on distribution. Therefore in this case, anything can happen with the two metrics.
The use of the z notes determines the difference recorded at the value of r. Compared to raw data, z-notes have several properties:
- the average of a distribution z is equal to 0
- the standard deviation of a z distribution is always 1
The use of z-notes in calculating the r-value is recommended when data have been collected using different scales.
The analysis performed by you only with raw data, the value of r is different because due to some errors, the distribution is not normal (extreme scores, multiple modes etc.).
To correct the errors, you can process the raw data as follows:
- elimination of multiple modes and preservation of a single-mode;
- using the average 5% trim which does not consider 5% of the scores at the extremities of the distribution.
After this processing, recheck the distribution to see if it meets the normality criteria.
Best regards,
- Badges
- Science topic
- Similar topics
- Linguistics
- Composition Studies
- Publication