“If you torture data for long enough, it will confess to anything.” Ronald Harry Coase
“Research can be statistically significant, but otherwise unimportant. Statistical significance means that data signifies something… not that it actually matters.” Paul Ingraham
A statistically significant research finding just tells you that some relationship may exist among the variables of interest, assuming the research was done properly (a big assumption). The relationship may be so teeny it’s trivial and relatively useless for prediction. For instance, let’s say a large study finds that Danes are “significantly” taller than Germans – by a whole half-inch on average. That doesn’t help anyone predict if any particular Dane or subset of Danes is taller than any particular German or subset of Germans. It doesn't even tell you if more Danes are taller than Germans or vice versa.
A lot of studies have such small sample sizes that findings of significance are pretty meaningless by themselves – they lack power. A lot of research is poorly designed and yields little in the way of useful data. Clinical trials that lack sufficiently controlled placebo conditions are especially prone to misleading assertions of significance.
Interventions, therapies, and treatments that involve personal contact may work their magic through the nature of the personal interactions rather than their unique special ingredients. For this reason, research on treatment effectiveness should include comparison conditions that control for personal factors such as how the researchers interact with subjects. That means the same amount of compassion, touching, attention, encouragement, and all-around support given to subjects across groups. Yeah, it's hard to pull off. But it's doable - if there's a will to do it. And then, if the study is large enough with the proper controls in place, any statistically significant finding may very well be... significant.