When medical stories in the mass media report dangers of certain agents (e.g., coffee drinking), later research often suggests that the effects may not exist or are weaker than first believed. Explain how this could happen if some journals tend to publish only statistically significant results.