+ Reply to Thread
Results 1 to 7 of 7

Thread: Interaction effect: only main effects significant

  1. #1
    Points: 13, Level: 1
    Level completed: 25%, Points required for next Level: 37

    Posts
    3
    Thanks
    1
    Thanked 0 Times in 0 Posts

    Interaction effect: only main effects significant




    Dear community,

    I'm running a regression model to test, if there is a difference between the Eastern and Western Germans supporting redistribution as a function of one's personal monthly income. In order to do this, I integrated an interaction effect between origin (0 = western Germany, 1 = eastern Germany) and the monthly income (metric).

    Both of the main effects have a significant effect, while the effect of income is negative and the effect of origin positive. The interaction effect has a non-significant (negative) effect. Do I have to interpret those results as follows:
    1. Among eastern Germans is no relationship between monthly income and support for redistribution.
    2. The higher the income of a person from western Germany, the lower her/his support for redistribution.
    3. People from eastern Germany have a higher support for redistribution than western Germans.
    ?

    Best,
    Marnie

  2. #2
    Points: 4,664, Level: 43
    Level completed: 57%, Points required for next Level: 86
    kiton's Avatar
    Location
    Corn field
    Posts
    234
    Thanks
    47
    Thanked 51 Times in 46 Posts

    Re: Interaction effect: only main effects significant

    Hello Marnie,

    Could you please specify how exactly you measure redistribution -- i.e., is it a binary variable (support/not) or is it some numerical value? In case of the former, you should run a logistic regression, whereas in case of the latter you can run a linear regression. The interpretation of the coefficients will be different for the two.

    Further, if you want to see if there are differences b/w Eastern and Western Germans in supporting the redistribution, you simply add your binary variable -- origin -- in the model and check if the coefficient is significant or not. If it is, then you should also calculate the effect size for the group difference. In this scenario income would act as a covariate (i.e., control).

    Otherwise, I don't really see a need to add an interaction. Having an interaction between origin and income would imply testing if income impacts redistribution depending on origin. For example, the magnitude of impact of income on redistribution is higher if you are a Western German and lower if you are an Eastern.

    Hope this helps

  3. The Following User Says Thank You to kiton For This Useful Post:

    Marnie (07-03-2016)

  4. #3
    Points: 13, Level: 1
    Level completed: 25%, Points required for next Level: 37

    Posts
    3
    Thanks
    1
    Thanked 0 Times in 0 Posts

    Re: Interaction effect: only main effects significant

    Hi Kiton,
    the dependen Variable is measured on a five-scale, where 5 indicates the highest support for redistribution.

    In your last paragraph you actually described better what I want to do than I did So, the main effect of "income" indicates the income effect size for indivduals whose value in the variable "origin" is zero, and therefore West Germans. And the main effect of "origin" indicates the effect size of being from the East (compared to West) when the income is zero (or on average, when you centered it in advance) ?

    But what does the effect size of "originXincome" tell me, when it's non-significant? That there is no relationship between income and support for redistribution among East Germans, I guess?

    Best,
    Marnie

  5. #4
    Points: 4,664, Level: 43
    Level completed: 57%, Points required for next Level: 86
    kiton's Avatar
    Location
    Corn field
    Posts
    234
    Thanks
    47
    Thanked 51 Times in 46 Posts

    Re: Interaction effect: only main effects significant

    When the interaction is non-significant, it points to the evidence that the affect of income on DV does not depend on origin. The effect if income on DV can still hold regardless of its interaction with origin. Now, when you will be interpreting the main effects note that in the presence of an interaction these effects become the so-called conditional effects -- i.e., say, the effect of income at the mean value of an origin. It is important.

    Also, since you have a 5-point scale DV it makes it actually an ordinal (or interval) type. While many in such case would still run a linear regression, methodologically it would be correct to run an ordered logistic regression. Just for your information.

  6. #5
    TS Contributor
    Points: 12,227, Level: 72
    Level completed: 45%, Points required for next Level: 223
    rogojel's Avatar
    Location
    I work in Europe, live in Hungary
    Posts
    1,470
    Thanks
    160
    Thanked 332 Times in 312 Posts

    Re: Interaction effect: only main effects significant

    Maybe a simpler picture: the plots of effect vs. IV are parallel lines, possibly shifted.

    regards

  7. #6
    Points: 13, Level: 1
    Level completed: 25%, Points required for next Level: 37

    Posts
    3
    Thanks
    1
    Thanked 0 Times in 0 Posts

    Re: Interaction effect: only main effects significant

    Quote Originally Posted by kiton View Post
    Now, when you will be interpreting the main effects note that in the presence of an interaction these effects become the so-called conditional effects -- i.e., say, the effect of income at the mean value of an origin. It is important.
    Is this also the case, when I dit not center the variable "origin"?

    I know that it would be better to run and ordered logistic regression, but since I'm using MLA it's not possible with my Stata-Version..

  8. #7
    Points: 4,664, Level: 43
    Level completed: 57%, Points required for next Level: 86
    kiton's Avatar
    Location
    Corn field
    Posts
    234
    Thanks
    47
    Thanked 51 Times in 46 Posts

    Re: Interaction effect: only main effects significant


    Centering has nothing to do with it. I personally like the following book a lot and recommend it for anyone exploring the interaction effects: Aiken, L. S., West, S. G., & Reno, R. R. (1991). Multiple regression: Testing and interpreting interactions. Sage.

    Are you sure your Stata version does not support -ologit- command? It should be supported by most versions.

+ Reply to Thread

           




Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts






Advertise on Talk Stats