If Fed decisions are data dependent, what do you do when the data is noisy and perhaps a little suspect?