I conducted a Longitudinal study on Iranian adolescents' physical activity. I am going to find mediators variables but am somewhat confused about running statistical analysis. Independent variable is the two groups according to stage of readiness (pre-adoption and adoption stages), potential mediators are related beliefs and outcome variable is the mean physical activity a day. To determine the mediators variables, is there a need to cancel any effect from independent variable to dependent variable when both hypothesized mediators and independent variables enter into regression analysis? Or if the relationship between independent variables and response is substantially reduced when both hypothesized mediators and independent variables enter into regression analysis? And then, to determine whether the reduction could be considered ‘substantial’, the Sobel test will be used.
Assuming by longitudinal study you mean something assessed on three or more time points, i don't think you should tackle this in regression. Potentially this could be handled as a growth model in HLM, but I think SEM would be the best choice. I suggest you start with a basis in Willett, JB & Sayer A (1994) for the basic growth model. There mediators become a question of "indirect effects." You could see that in my papers "Does child care involvement increase or decrease distress among dual-earner couples? " and "Trajectories of Internalizing Problems in War-Affected Sierra Leonean Youth: Examining Conflict and Postconflict Factors." Bob
Dear Robert, I have the same problem and I want to do mediation analysis in AMOS (one categorical predictor, continuous mediator and criterion is also continuous, but it is set as repeated measures in 4 times, with latent factor for intercept and slope), but I am confused how to set mediation - does mediator should be related to slope or intercept latent factor or somewhat different? Is there any manual how to do it in AMOS? Thank you in advance.
- Badges
- Science topic