-- BarbaraGrimes - 29 May 2009

Checklist for Corr Macro

Check the value of n=number of number of subjects with non missing values for both var1 and var2. Is this the number that you expected?

Are the Spearman and Pearson correlations similar? If not, the Spearman correlation (also called "rank correlation") may be a more valid summary, because it cannot be as strongly influenced by a single extreme point and can better reflect relationships that are monotonic but not linear.

For any relationships of particular interest, examining a scatterplot is recommended. Note that two variables can have a strong but non-monotonic relationship even though Pearson and Spearman correlations are both close to zero.

Correlation is a unitless, abstract quantity that may be hard to interpret. Consider regression of one variable on the other as a complementary way of further describing the relationship.