您可以像这样显式查询“反弹”的数量：

select count(*)
from t as t1
where t1.event = 'visit'
  and not exists (select * from t as t2 where t1.received < t2.received and t1.sessionid = t2.sessionid)

不确定“反弹率”的分母具体是什么？每次反弹？#反弹/#事件？

嗯。我想你应该总结一下 session_id 并计算访问的类型。然后聚合。你的问题不是100%清楚，但我认为：

select (count(*) filter (where num_notvisits = 0) * 1.0 / count(*)) as bounce_rate
from (select session_id,
             count(*) filter (where event = 'visit') as num_visits,
             count(*) filter (where event <> 'visit') as num_notvisits
      from t
      group by session_id
     ) s
where num_visits > 0;

这是访问和非访问事件的会话数除以访问会话数的比率。
实际上，您可以将“外部选择”更简单地表述为：

select avg( (num_notvisits = 0)::int ) as bounce_rate