The method provides for the fitting of a different distribution to the tail of a distribution of continuous data, than that distribution fitting to the rest of the distribution. By fitting a distribution to this set of occurrences only, better distributions shapes are obtained for that part of a distribution which might otherwise be overly sensitive to individual or small numbers of occurrences. Over sensitivity can make such distributions unreliable in situations where their value is compared with the value of another distribution. A distribution which does not decline quickly to zero is preferred for the tail, for instance a heavy tailed distribution.