Interaction

[Mindy Green]

It turns out that the two variables are not significant, but after adding the interaction of these two variables, the two variables and the interaction item are significant, what does this mean?

[Richard Williams]

It can happen. For example, the effect of income might be positive for men but negative for women. In the model with no interactions, the positive and negative effects might cancel out. But, once you add the interaction, the effects could become significant.

Lots of other things could be happening too. For example, the estimated effects between the 2 models could barely differ but in one the coefficients could be barely significant while in the other they could be barely insignificant. It isn't like P values are absolute truth statements.

We can't tell what is going on in your case, but there isn't any inherent reason that makes results like yours impossible to occur.

[Mindy Green]

Dear Richard Williams ,

Thanks so much for your interpretation. It helps me a lot!

Best wishes