> It should be strong if the effect is strong. We find strong signal in noisy data in all kinds of contexts

And we fail to find signal, or can find whatever "signal" we want to find in all kinds of other noisy data. What says any kind of useful info is here to find?