A very large survey finds a statistically significant association between age and support for a policy, but the estimated difference is only 0.2 percentage points. What should the researcher avoid concluding from significance alone?