1
$\begingroup$

Please excuse if this question has already been asked. Is it possible to include several interaction terms into one model (I am keeping the main effects in). So I have:

M1: X1, X2, X3

M2: X1, X2, X3, X1 x X2

M3: X1, X2, X3, X1 x X3

M4: X1, X2, X3, X1 x X3, X1 x X2 (this one gives me results very hard to interpret - due to strange sign changes)

I would like to go with only M1-M3 but I am worried that reviewers will ask why I did not do M4.

Improve this question
$\endgroup$
2
  • 1
    $\begingroup$ Yes, it is possible to do it. I could indeed see that a reviewer might ask that question. Is it perhaps easier to interpret the predictions for particular covariate combinations? $\endgroup$
    – Björn
    Commented Jan 3, 2016 at 10:49
  • $\begingroup$ Thank you for your answer. I am mainly struggeling with the interpretation that in M2 and M3 the interaction terms are both negative and when I put them together in the model one of them is either not significant or there is a positive significant interaction term (M4). I am just wondering if there is an additional value in including them both. Thank you in advance $\endgroup$
    – Laura Weis
    Commented Jan 3, 2016 at 10:56

1 Answer 1

Reset to default
3
$\begingroup$

First, it is possible to have multiple interaction terms in a model.

Second, people may ask why you included some interaction terms and not others - you need to have a good answer.

Third, "things were hard to interpret" is not a good answer.

So, how to look at model M4? In addition to trying to interpret the parameters, you should look at the predictions the model makes. Look at the predicted Y for various combinations of X1, X2 and X3. Make graphs. What graphs you should make depends on the nature of the X variables but if all are continuous then coplots can be good.

Improve this answer
$\endgroup$
1
  • $\begingroup$ Thank you for your helpful answer. I am working with social network data, therefore plotting parameters is difficult. I know which parameters I would like to include, the main question I had (and was answered) was if it will give me meaningful results in general. Thanks again $\endgroup$
    – Laura Weis
    Commented Jan 3, 2016 at 14:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.