Not if you want to claim statistical significance. The math behind this method is based on defining the hypothesis before seeing the data (and even then it's usually very weak evidence of a tiny signal within the noise).