Hello dear researchers
In my study, I have 5 independent variables and 1 dependent variable. In multiple regression analysis, not all my IVs significantly predicted my DV. However, most of my IVs predict the participants' age and BMI value; in turn, age and BMI value significantly predict my outcome variable. Does that mean a mediation or an indirect relationship? Also, should I continue with SEM or just a path analysis?
In terms of causal ordering, I find it difficult to conceive of individuals' age (e.g., in years) as a mediator (endogenous) variable. What exogeneous variable would cause age? Before running an analysis, I would recommend reading some good resources on causality and on statistical mediation analysis (there are plenty of relevant didactic papers and books out there!). For example:
https://www.amazon.com/Correlation-Causality-David-Kenny/dp/0471024392
Article An Introduction to Statistical Mediation Analysis
MacKinnon, D. P. (2008). Introduction to statistical mediation analysis. Taylor & Francis Group/Lawrence Erlbaum Associates.
https://www.guilford.com/books/Introduction-to-Mediation-Moderation-and-Conditional-Process-Analysis/Andrew-Hayes/9781462549030
Mediation analysis with observed variables is most easily conducted through path analysis, which allows you to estimate all paths simultaneously and to test the indirect effects for statistical significance. Latent path analysis (SEM) can be conducted if you have measurement models with multiple indicators for some or all variables in your conceptual model.
Hi Ayşe Dikyol Mutlu ;
We cannot know whether there is a mediation relationship unless we test it.
If you think that there will be a mediation relationship between these variables in the literature and logically, you can perform a mediation analysis.
You can use SEM if your variables are obtained from scales (if your sample size is sufficient).
Bests
Thank you, Christian Geiser and Abdullah Faruk Kılıç for your informative answers,
I will consider the links in detail. You are right for sure about the mediator role of age. Considering its relationship with both the predictor variables and the outcome variable, it would probably be better to test its moderator role in the relationship. Yet, if X and Z(in my example) do not have a direct relationship, can I still go for the testing of moderation?
Yes, you can still conduct the moderation analysis. Macckinnon (i guess) stated that although there is no relationship between X and Z, the moderator effect can be examined.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression