The problem is very similar to the previous section, but we consider words of rank >N instead.
For third table line:
select 100,(select 1+count(distinct(w2_id)) from co_n where w2_id>200) / (select 1+count(distinct(w1_id)) from co_n where w1_id>200);
5.1.9.3 Skewness in NN co-occurrences III