I am now analysing the suicide data to mine for bunching and identify possible groups of interest. Of course this presents the statistical problems of cherry-picking and post-hoc hypothesising, but it may reveal something worthwhile.