I think if you measure random unrelated data sets against each other you're quite likely to find spurious correlations. If you design an experiment between pote

I think if you measure random unrelated data sets against each other you're quite likely to find spurious correlations. If you design an experiment between potentially related variables with a potential hypothesis on how they are related and then observe data that matches the hypothesis, you probably have a reasonable chance the hypothesis is actually the explanation.