Interactions in regression analysis

#1
Is their anyone who knows how to identify interactions between independent variables in a regression analysis? How does one use the output diagnostics to determine that there is an interaction.

Thanks
 
#2
You can look at pairwise correlations between the independent variables, if r values are greater than 0.80 (or smaller than -0.80) the variables are strongly correlated.
 
#3
Hi Quark

Thank you for your response. From what you say, I infer that if two variables correlate with one another then they will interact? I have not come across this in my statistical studies. If two variables are highly correlated I am inclined to consider removing one from the analysis rather than try to model an interaction between them.

But I know that interactions are possible between variables that are not correlated; how would I detect interaction effects between such variables?
 
#4
But I know that interactions are possible between variables that are not correlated; how would I detect interaction effects between such variables?
There's a VIF(Variance Inflation Factor), which measures the impact of collinearity among the X's in a regression model. Most statistics software will show it in the multiple regression output. You can google it to learn more.
 
#5
Hi Quark

Two points here.

1) Does correlation imply interactions? your responses so far seem to imply so, but I have never heard of this before.

2) Yes, VIF has a lot to do with collinearity, but how is it related to interaction effects, especially when the variables are not correlated as mentioned in your quote from my previous mail
 
#6
1) In multiple regression context, if two IVs are significantly correlated, then chances are there are significant interactions. There's no need for a broad statement like "correlation imply interactions".

2) Multicollinearity is interaction between IVs. Typically a VIF value greater than 10 MAY suggest multicollinearity.