 Mediation ja moderation models are very common in social sciences, and they are often combined, and there is a large body of current literature that discusses the combination or whether you use mediation or moderation. The idea of mediation was that when you have x and y, then the effect of x and y goes to a third variable called the mediator. The idea of moderation was that when you have a relationship between x and y, then the strength of that relationship varies as a function of a third variable, which here is the z. So these two effects are sometimes confused, so it may be difficult to make the choice of whether your effect is actually a moderation effect or a mediation effect. The Baron Kehney 1986 article was written to address these concerns that research are confusing these two models, so which one is appropriate for the data. It is less common nowadays, but one thing that you should know is that if you are choosing between mediation model and a moderation model, you don't have to because you can do also moderating mediation. So you can combine these variables together, these models together, so you can have a mediation model and then you can have a moderation effect for all of these three paths. You could also have a m here moderating this direct effect for example, so you could have a m moderating itself in which case you would have a u shape effect for this relationship. So it really gives you lots of options for variables, how you model them. And then you have to make a choice based on your theory, what is the best representation of your theoretical ideas, so what's the best model for your particle study. There is yet another concern that you need to take into account. The concern is that the mediation model is just identified. If you just look at correlations, then the mediation model will always explain those correlations perfectly. So it can be tested empirically just with three variables. Another concern is that just by running a regression of x and m and y, you can run the same mediation model in six different ways just reversing some of the regression paths. So this is the same model with just the positions of x, m and y reversed or altered. So you can't really say which one of those is correct empirically because they all produce the exact same model implied correlation matrix. This is from Klein's paper, The Mediation Myth. And his main argument or one of the main arguments is that if you do mediation, you can really only say that there's a mediation effect if you have long internal data sets. So you can't from a cross-sectional model data to say that this model is the correct one and not that one, because these both fit the same data equally well. Similarly, it is possible that instead of this model, you have this model with a bidirectional effect. So this is called a simultaneity in econometrics. We don't address that now. And it's also possible that x is just a common cause of m and y, and m and y are correlated for some other reason, but there is no direct effect between m and y. So empirically, all of these models are equally good. So you have 18 models, all of them are plausible. So which one is the best for you? That's where your theory comes to play. So you have to theoretically argue unless you have long internal data. You have to theoretically argue that your model is one of these or one of these or one of these and not the other one. And that can be difficult to do. Mediation testing is an active topic. A mediation model estimation is an active topic. So every year there are more new papers and more techniques coming. And the practices in mediation and moderation testing are changing. So one good way for mediation is to go to David Kenny's website. And that maintains a list of up-to-date sources for some of the reason discussions in mediation. So should you use Baron and Kenny, should you use bootstrapping or should you use the Sobel test? Different perspectives on these problems exist. And David Kenny's website contains a good list of references for some of the most central articles. Here are some and also if you do a mediation model, I recommend that you search organizational research methods for the term mediation and see what is the current practice in organizational research. There are at least one or two articles that come out each year that address mediation because this is such an important topic.