Hi all, I have a question about mediation analysis.
In my design, subjects were asked about their relative preference between two options, i.e., to what extent they would prefer option B relative to option A. For example, option B has a cuter design than option A. My research question is whether a cute design would improve the product's attractiveness. So the IV actually has two levels (cuter design vs. less cute design), but they were both presented to subjects, and could not be divided according to the data... I'm not sure whether it should be called one level or two levels.
To simplify the question, the IV is whether the product is cute or not, the DV is the relative preference between a cuter product and a less cute product. The proposed mediator is relative attention assigned between the two products.
I think I could perform a one-sample T-test to test the impact of cute design on preference.
But how could I perform the mediation analysis?
Thanks a lot!
Li Zhang In principle, path analysis is most convenient for estimating and testing mediated (indirect) variable effects, as it allows you to simultaneously consider more than one outcome (dependent) variable (i.e., estimating two or more linear regression equations simultaneously) and estimate/test the indirect effect as part of the analysis. Path analysis can be done in programs for structural equation modeling such as lavaan (in R; free software), Mplus, AMOS, EQS, LISREL, and OpenMx.
However, I'm not sure path or mediation analysis would work for your design which seems to be a within-subjects design (you wrote that all subjects rated both options)--if I understand correctly. A typical mediation analysis would imply that you have two (or more) groups (X, ideally, randomized to cute vs. non-cute stimuli) for which you then measure the outcome (Y) and the mediator (M). In a between-subjects design like this where your IV (X) is a grouping variable (independent and ideally randomized experimental groups), you could examine the indirect effect X --> M --> Y. Unless I'm missing something, I'm afraid with your current within-subjects design (where X cannot be expressed as a between-subjects/grouping variable), this isn't possible to do.
@Li It seems that your IV is binary while you have not mentioned the type of data for your DV and Mediator. In this case, you can use PLS-SEM in Smart-PLS as it works perfectly fine with categorical/binary variables.
- Badges
- Science topic
- Similar topics
- Cognitive Science
- Thinking