Proponents of this approach insist it is the only way to tell exactly what the effect of the change is.
If you already have some idea of the kinds of interactions there are among the variables (or if you know you have no idea), you can design an experiment so as to determine fairly precisely which variables are important, and which are not. Such an experiment design can even tell you what the interactions are between variables. Such design of experiments is not a trivial task. Handbooks of worked-out designs make it easier.
In the real world, there are almost always interactions between variables, and those interactions are not necessarily linear. (More of an input does not necessarily produce more of an effect.)
It has been pretty well established in a variety of experiments that the experimenter's conscious expectation of results can actually cause the expected result. Even blind or double-blind experiments can suffer from problems in this area. Where consciousness is involved, exceptionally careful experiment design is required so that the result is not affected by someone just thinking about what s/he expects. [Do not confuse this with the bias of an experimenter interpreting results. That is a problem, too, but of a different kind.]
If you already have some idea of the kinds of interactions there are among the variables (or if you know you have no idea), you can design an experiment so as to determine fairly precisely which variables are important, and which are not. Such an experiment design can even tell you what the interactions are between variables. Such design of experiments is not a trivial task. Handbooks of worked-out designs make it easier.
In the real world, there are almost always interactions between variables, and those interactions are not necessarily linear. (More of an input does not necessarily produce more of an effect.)
It has been pretty well established in a variety of experiments that the experimenter's conscious expectation of results can actually cause the expected result. Even blind or double-blind experiments can suffer from problems in this area. Where consciousness is involved, exceptionally careful experiment design is required so that the result is not affected by someone just thinking about what s/he expects. [Do not confuse this with the bias of an experimenter interpreting results. That is a problem, too, but of a different kind.]
