Question

我正在尝试以假设为基础进行计算，所以我假设我可以使用函数的情况完成此操作。如：

SELECT                                                                                                          
TRIM(EXTRACT(YEAR FROM business_date))||'Q'||TRIM((EXTRACT(MONTH FROM business_date)+2)/3) as year_qtr,                 
case when orr_txt = '1' then  (sum(cast (rr9m as float)) / (sum(cast (total_observations as float)) - 0.5*sum(cast (lft as float)) )) end as  DR_1,
case when orr_txt = '2' then  (sum(cast (rr9m as float)) / (sum(cast (total_observations as float)) - 0.5*sum(cast (lft as float)) )) end as  DR_2,
case when orr_txt = '3' then  (sum(cast (rr9m as float)) / (sum(cast (total_observations as float)) - 0.5*sum(cast (lft as float)) )) end as  DR_3,
avg(cast (dp as float)) as Pred_PD                                  

FROM                                        
     raroc.pd_matrixRR a,                                       
     raroc.ccc_dp_sc  b,
     raroc.ratings_convert c                                        
WHERE                                                                                                   
     start_orr_code in (1)
     and case when a.scorecard_group = 'BB' then 'Business Banking' else a.scorecard_group end= b.rating_sys
     and a.orr_txt = b.bacrr
     and a.orr_txt = c.bacrr_text
     and orr_txt not in ('9-', '10', '11', '12', 'NR')                                      
     and rating_sys_grp = 'Scorecard'
     --and a.orr_txt = '1'
     --and scorecard_group = 'Large Corporation'
     --and scorecard_group = 'Large Corporation_FI'
     --and scorecard_group = 'Middle Market'
     --and scorecard_group = 'BB'
     --and scorecard_group = 'GCSBB'
     --and scorecard_group = 'Real Estate'
      and scorecard_group = 'Individuals'                                       
      and scorecard_group not like 'Sove%'                                      
      and scorecard_group not like 'Lega%'
      --and year_qtr < '2014Q2'                                                                                         
group by 1,2,3,4                            
ORDER BY 2,1

显然没有编译并嘲笑我的逻辑。但是，如果我不对它进行分组，我将收到错误3604。问题是，我如何得到这个来计算第1,2和3类的比率。非常感谢！

SP

Answer 1

我认为你只想要条件聚合。条件 inside 聚合函数。我认为这是你想要的表达式（当然每个值重复3次：

SELECT . . .
       (sum(case when orr_txt = '1' then cast(rr9m as float) end) /
        sum(case when orr_txt = '1' then cast(total_observations as float) end) -
        0.5 * sum(case when orr_txt = '1' then cast(lft as float) end)
       ) as DR_1, . . .

然后，您可以将GROUP BY更改为GROUP BY 1。

错误3625：分组依据，可能不包含聚合函数

1 个答案: