Comment by danwiding

The behavior of the top decile is confusing as it's laid out presently. If the it's possible that the top 99th percentile drinks on average 16 drinks per night which would make it possible for the 91st decile drinking only 4 drinks per night and still get the results shown in the graph. My bet is that the data here is getting heavily skewed by a subset of alcoholics at the top end.