Predictive designs and analyses can get messy. We observe the relationship between two variables, often witout taking in to account the noncontrolled nature of such a design, and we learn out the fact that many other variables may be at play, affecting the relationship between those two variables. Sometime researchers take this in to account, and sometime they don't. Please provide a real or made up example of when predictive design and analyais may be problematic or even detrimental if researchers do not consider other variables at play.