“…claims involves chaining them together "piece by piece," for example, by running multiple experiments. Combining such independent estimates still requires substantial assumptions (e.g., Imai et al, 2011, p. 770), although they may often be more defensible (Pirlott & MacKinnon, 2016;Strobl & Wunsch, 2018). At the same time, this of course presumes that the mediator can be manipulated in a targeted manner, which may not always be the case, in particular if it is a psychological variable (Eronen, 2020).…”