After completing the simple and multiple regression sections, I was feeling fairly confident with how I was doing. Perhaps it was a good thing then that the next two topics, moderation and mediation, snapped me back to reality.
It was when I went to start the quiz, that I realized I did not have a firm grasp on the differences between moderating and mediating variables. Almost every question was a struggle to figure out what was even being asked, so I went in search of clearer definitions and additional practice questions.
Moderation
Field describes moderation as the relationship between two variables that change as a function of a third variable. In other words, a third variable moderates the effect of the predictor on the outcome.
In statistical terms, this is known as an interaction effect. It is only by adding an interaction term to the regression equation that moderation effect can be detected.
When looking at a moderation model, there are three terms that determine the outcome variable: predictor, moderator and predictor*moderator (the interaction effect). If the interaction effect is close to zero, then there is no moderating effect. However, if the interaction effect is significant, then there is a moderating effect.
Additional information on moderation: https://edge.sagepub.com/field5e2/chapter-specific-resources/11/cramming-sams-top-tips
Centering Variables
If a moderator, or interaction term, is present then the variables need to be centered. This makes the regression coefficient easier to read and interpret.
In this case, grand mean centering is used. It transforms the values of a variable into deviations around the mean of the variable. In other words, the interpretation is no longer about zero, but about the group average. For example, X – X-bar = the deviation of X.
Coding Variables
Predictor variables in regression models can be continuous or dichotomous. For dichotomous variables, (k – 1) variables represent the categorical variable and they can be coded in three ways: dummy, effect or orthogonal coding.
Additional information on coding: https://stats.idre.ucla.edu/spss/faq/coding-systems-for-categorical-variables-in-regression-analysis/
Mediation
Field defines mediation as the relationship between a predictor variable and outcome variable when the strength of the relationship is explained (reduced) by a third variable.
It is important to note that partially mediated relationships are more common in practice than fully mediated relationships.
The mediation effect cannot rule out other variables from effecting the outcome variable (i.e., cannot show causality), but it can help evaluate whether a mediated model is a plausible hypothesis. If mediation is perfectly in effect, then the relationship between the predictor and outcome variable would be completely gone once the mediator is added to the model.
Additional information on mediation: https://edge.sagepub.com/field5e2/chapter-specific-resources/11/cramming-sams-top-tips
Four Conditions of Mediation as proposed by Baron and Kenny (1986)
The four conditions of mediation were proposed by Baron and Kenny (1986):
- The predictor must significantly predict the outcome variable (i.e., the first model, c must be significant).
- The predictor must significantly predict the mediator (i.e., the second model, a must be significant).
- The mediator must significantly predict the outcome variable (i.e., the third model, b must be significant).
- The predictor variable must predict the outcome variable less strongly in model three than in model one (i.e., c’ must be less than c or c’<c).
Refer to the Mediation Model above for visual depiction of a, b, c and c’.
What’s the Difference?
This was what I struggled with the most. I understood the difference between moderators and mediators, but when I was provided a scenario I was having difficulty keeping those differences clear in my mind.
The clearest definition I found was provided by Pritha Bhandari (https://www.scribbr.com/methodology/mediator-vs-moderator/). Bhandari states, “a mediator [is] a go-between for two variables,” while “a moderator is something that acts upon the relationship between two variables and changes its direction or strength” (2021). It was only after I read and re-read the definitions and examples Bhandari provided that I was able to answer practice question involving moderators and mediators with some accuracy.
References
Anderson, D. R., Sweeney, D. J., Williams, T. A., Camm, J D., & Cochran, J. J. (2018). Statistics for Business & Economics (13th ed., revised). Cengage Learning.
Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51(6), 1173–1182.
Bhandari, P. (2021, September 15). Mediator vs moderator variables. Scribbr. Retrieved September 23, 2021, from https://www.scribbr.com/methodology/mediator-vs-moderator/.
Field, A. (2017). Discovering statistics using IBM SPSS statistics (5th ed.). SAGE Publications.
Howell, D. C. (2017). Fundamental Statistics for the Behavioral Sciences (9th ed.). Cengage Learning.