An addendum to that is that that if a sufficiently large and diverse dataset is interrogated, then one can find connections between person A and criminal B, when such connections, while real, are accidental.