QM LM9 Parametric and Non-Parametric Tests of Independence

Question 1

What is one problem with testing for significance of correlations in a large data set?

Answer 1

• In a large data set even minor relationships can have significance
• Meaning a researcher could simply increase the size of their dataset to find significant correlations (a kind of cherry picking)

Question 2

When do we use a non-parametric correlation test?

Answer 2

• If normality assumptions for x or y are violated (non-normal distributions) or
• outliers are present
• we use a spearman rank correlation coefficient
• which is basically a correlation test but between ranks rather than values