A study's overall estimate is precise, but an estimate for a small subgroup has a wide confidence interval. What is the most likely explanation?