# Significant interaction effect?

#### jesse

##### New Member
I would like to test how how gender interacts with another (continuous) variable. To do this, I am using a hierarchical (nested) regression model.

The first model uses the continuous variable and gender as predictors, but only the continuous variable is statistically significant.

In the second model, the continuous variable, gender, and the multiplicative interaction term are included as predictors, and all of them have a statistically significant effect.

Are the statistically significant relationships in the 2nd [fixed typo] model validly so? Why is gender non-significant in the first model, but significant in the second?

Last edited:

#### hlsmith

##### Less is more. Stay pure. Stay poor.
Can you describe what you are calling a hierarchical model? Your description just seems like multiple models. What are all of the variables including dependent variable.

#### spunky

##### Can't make spagetti
Can you describe what you are calling a hierarchical model?
it means the OP is adding predictors in blocks to construct larger models where the smallest ones are nested versions of the immediate larger ones. tradition says that if you follow this process, assess the significance of the R-squared-change and it turns out to be significant, that block of variables contributes to the explanatory power of your regression model and should be retained.

#### hlsmith

##### Less is more. Stay pure. Stay poor.
What is your third model the one with the interaction term?

List out all of your models out, the post is written in a way that is confusing which variables are in which model.

#### jesse

##### New Member
@spunky: Thank you for clearing that up.

@hismith: Rereading the post, it is not only confusing, but incorrect. There is no third model - only two models.

Model 1:
y = x1 + gender
Model 2:
y = x1 + gender + x1*gender

In model 1, only x1 has a statistically significant effect at .05. However, in model 2, both independent variables have a statistically significant effect, as does the interaction term (x1*gender).

My question is - is the statistically significant effect of x1*gender in model 2 a legitimate result, even though gender was not significant in model 1?