confounders

Summary

Third variable C associated with both A and B, A may not actually be associated with B

Concept

Reduced by random sampling(distributing characteristics that may be confounders evenly between both groups) or slicing(in obervational studies)

Simpson’s Paradox

• Trend/association between 2 variables appears in more than half of the sub-groups, but disappears or reverses when the groups are combined
• Proves the existence of a counfounder