How to interpret mediator effects on response and vice versa? just want to get my head around mediator and how to properly interpret its effect on the relationship between predictor and response.
From https://journals.lww.com/jnpt/fulltext/2019/04000/mediators_and_moderators,_confounders_and.1.aspx, the authors explained that the mediators "lie on causal pathway" whereas the covariates and confounders "do not lie on the causal pathway". As such, it kind of mean that there's no way to tease the mediators from the predictor if the mediators significantly contribute to the response, i.e. there's some "interaction" between the predictor and mediators which leads to a causal effect on the response? Is this correct?
To help the above discussion, let's say the mediator is overweight and the response is hypertension.
Now if I ask a different question such that the mediator interchange with the response (overweight becomes response and hypertension becomes mediator), and the association is significant, this then means that there's an interaction between the predictor with hypertension (formerly analysed as response) causing an effect on overweight (formerly analysed as mediator). Consequently, doesn't this mean there's an interaction between the mediator and response since significance will be observed irrespective to if they are treated as mediator or response? And would be incorrect to say overweight is mediating the effect of predictor on hypertension because we can also say that hypertension is mediating the effect of predictor on overweight?
I'm very confused, can someone help to shed light into this darkness? Thanking in advance.
 A: One says that there is an 'interaction' between the moderator and the predictor, not between the mediator and the predictor, as you are doing.
To make the example of a mediator let us assume that:
Obesity (the predictor) increases the risk for diabetes (what you call the response).
Now you might ask, what is the mechanism underlying this (supposed) causal relationship? One hypothesis is that insulin resistance mediates this relationship, that is, is the mediator. Obesity increases insulin resistance, which in turn increases the risk for diabetes. There needs to be no interaction between obesity and insulin resistance. Obesity remains a predictor both for insulin resistance and diabetes. Insulin resistance might provide a deeper/biological understanding of how/why obesity increases the risk for diabetes.
In this context, an example for a moderator could be genes.
There might be some specific genes, such that if one person has them, despite being obese the risk for diabetes is not increased. In this case, one speaks of an interaction between obesity (the predictor) and genes (the moderator).
Hope this helps.
