I would like to test whether an intervention boosts self-efficacy at time 1, which then lasts through self-efficacy at time 2, and at time 3, etc, to then have an impact on medication adherence at time 5. Is serial mediation an appropriate method for this question? I would be using the PROCESS macro.
$\begingroup$
$\endgroup$
1
-
$\begingroup$ Are you interested in the effect of the intervention, or the extent to which self-effiacy persists? Because you could simply directly estimate the effect of the intervention on medication adherence and on self efficacy at time 5, 4, 3, 2, 1. Using IV, you could also estimate the effect of self-efficacy on medication adherence, under the assumption that the intervention can only affect it through self-effiacy. All of this assumes that the causal effect of the intervention on the other variables is identified, perhaps because you have controlled its provision in an experiment. $\endgroup$– CloseToCMay 26, 2021 at 12:52
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
First, yes it can. You will be in a slightly different framework though, which is longitudinal mediation. It would be consider a cross lagged mediation panel, but a weak one, given the design presented (a single measure of adherence?).
My two cents on your analysis : if your design enable it and are willing to use structural equation modelling, I would consider a latent growth model of self-efficacy, from which I would use the latent variable as mediators. It seems much better and appropriate to answer your question.
