1

I conducted an ANOVA for a 2x2 design and obtained a main effect and an interactions. However, I also measured a variable that is known to impact this effect and I re-did the analysis in an ANCOVA. This time the main effect disappeared, but the interaction remained significant.

What does this mean for the validity of my effect?

Apologies if this is a basic or banal question.

kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467
Emir Efendic
  • 31
  • 3
  • *Short answer: a certain effect could **correlate with both cross terms and main terms**, such that changing the model, by adding or changing a cross term, may remove the significance of an earlier found significant main term.* Long answer: Could you clarify whether you mean that 'you changed one variable from categorical to continuous', or 'you added a continuous variable'? Also, a more specific example of the data or problem would be helpful. – Sextus Empiricus Nov 06 '17 at 10:58
  • https://stats.stackexchange.com/questions/5450/what-if-interaction-wipes-out-my-direct-effects-in-regression – Sextus Empiricus Nov 06 '17 at 11:01

0 Answers0