The dose is the key. Too much of it and you are likely to be making lots of Type I errors—false positives—in which you find connections that are not really there. Too little and you make Type II errors—false negatives—in which you miss connections that are real. The signal-to-noise ratio is everything.
