8.3 Identifying a moderator

Here are 3 scenarios demonstrating how a third variable can modify the relationship between the original two variables.

• Significant –> Non-Significant
  • Significant relationship at bivariate level
  • We expect the effect to exist in the entire population
  • Within at least one level of the third variable the strength of the relationship changes
  • P-value is no longer significant within at least one subgroup
• Non-Significant –> Significant
  • Non-significant relationship at bivariate level
  • We do not expect the effect to exist in the entire population
  • Within at least one level of the third variable the relationship becomes significant
  • P-value is now significant within at least one subgroup
• Change in Direction of Association
  • Significant relationship at bivariate level
  • We expect the effect to exist in the entire population
  • Within at least one level of the third variable the direction of the relationship changes
  • Means change order, positive to negative correlation etc.