Could the effect be at least in part explained as a statistical fluke, like cancer clusters that can look ominous but turn out to be random?

Chance may play some role in situations like this, said Donald Berry, a statistician at MD Anderson Cancer center. What matters, he says, is not the number of cases but the clusters.

Suppose an executive on Wall Street was infected early in the epidemic and exposed a group of others by shaking hands. It was the start of a cluster that ballooned out as each person infected others. Suppose that at the same time an actor on Broadway was infected and started another cluster. And a patient in a nursing home started yet another cluster. The number of cases can start to explode.

Analyzing spread is not a case by case situation, Dr. Berry said. It is a cluster by cluster issue.