So im investigating whether psychological closeness (independable variable) and stigma by association (mediator) influence whistleblowing intentions (dependent variable). I also measure if honesty-humility (moderator) influences this relationship. So I have a moderated mediation. Respondents were either psychologically close with the wrongdoer or psychologically distant, and there either was stigma by association or there was no stigma by association present, resulting in a 2x2 design (4 hypothetical scenarios to which respondents were randomly allocated). In each of the 4 scenarios, they were asked how likely they were to blow the whistle on the wrongdoer (1-5 likert scale). After this, their honesty-humility was determined by asking 10 questions (1-5 likert scale). How would I statistically test this model?
*note: I did not measure the impact of psychological closeness on stigma by association. They were either present, or not present.
Hello Christof,
Path analysis is a common way of evaluating models such as the one you describe. Do note that, the moderating effect can be proposed to act on the IV-mediator relationship, or on the mediator-DV relationship.
Here's a link that explains the approach pretty clearly: https://scholarworks.umass.edu/cgi/viewcontent.cgi?article=1404&context=pare
The fact that both the IV and proposed moderator are dichotomous has no impact on using path/regression estimation. The real question has to do with whether you consider your intent variable to be of interval strength or just ordinal strength, as is the case also for the proposed moderator, given that these are based on responses to one or more Likert-type response scales. If you ultimately decide that ordinal is the better characterization, then you'll need to analyze via a method which allows for that. Programs such as the commercial Mplus, or the freely available R system library, lavaan, are two examples which could be used.
Good luck with your work.
- Badges
- Science topic
- Similar topics
- Social Psychology
- Attitude