Good Day!
I am now conducting a quantitative study utilizing a descriptive correlational methodology. I am also employing an Independent Variable and Dependent Variable with a Moderating Variable as the third variable. Can you assist me to determine what statistical treatment is essential to investigate the correlation of the IV towards the DV with the intervention of the MV?
Thank you very much!
To test moderation, you need to assess whether there is an interaction between the IV and the moderating variable. You can do this with a regression that includes the IV, the MV, and the product of the two, IV*MV.
Ella De Jesus
As mentioned by David L Morgan , the moderated effect can be tested in a regression model by including a product term (moderator*IV) in the regression equation. This is called moderated regression analysis.
Mathematically, a moderated regression equation for one IV and one moderator variable looks like this:
Y = b0 + b1*X + b2*Z + b3*X*Z + error
where Y = dependent variable, X = independent variable, Z = moderator variable, and b0, b1, b2, and b3 indicate constant regression coefficients. In order to test for moderation, you multiply the X and Z variables to obtain a product term (X*Z), for example, by using the COMPUTE command in SPSS. The product term X*Z gets added to your model as an additional "predictor." The moderator Z can be binary (a 0-1 coded dummy variable) or continuous (metrical/interval scale). In case of continuous X and Z, it is often useful to center X and Z before computing the product term to facilitate the interpretation of the results (in particular when the uncentered X and Z have no meaningful zero points).
The coefficient b3 for the interaction (product) term allows you to examine whether there is an interaction (moderator) effect. The null hypothesis H0 for the interaction effect is H0: b3 = 0. That is, when b3 = 0 in the population, there is no moderation (interaction) effect. When b3 is significantly different from zero, this null hypothesis can be rejected.
For details, see Aiken and West (1991):
Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and interpreting interactions. Sage Publications, Inc.
David L Morgan Christian Geiser
I appreciate your responses. Thank you so much!
How about the DV? Is the DV included in the multiple regression analysis?
I am just confuse right now. Thank you for enlightening me :)
@Jesus Indeed your DV has to be included in the regression model as outcome variable (Y). @Christian has already given you a detailed step by step procedure to check for moderation/interaction effect using multiple linear regression model. However, you could also use model-1 in process macro for SPSS to check for simple moderation.
By definition, the dependent variable is always included in a regression analysis. In the equation provided by Christian Geiser, it is labelled Y.
If your sample size is adequate you might also consider SEM which has the advantage of controlling for measurement error.
- Badges
- Science method
- Similar topics
- Methodology