SAS Proc sql中的此代码有什么问题

Question

I've tried to run the following code but it gave me an error message. Please see the following.

%macro bktest(prod=);
proc sql;
create table &prod._pd_coh_seg as   
select distinct
ASOF_YYYYMM,
segment,
mean(rcpa_b2_pd_low),
mean(rcpa_b2_pd_high),
sum(rcpa_weight) as Count,
sum(annual_default_occurrence*rcpa_weight)/sum(rcpa_weight) as dr_ct,
sum(dp_pit*rcpa_weight)/sum(rcpa_weight) as pd_ct_pit,
sum(RCPA_B2_PD_TTC_SEG_FCTR*rcpa_weight)/sum(rcpa_weight) as pd_ct_ttc
from backtest.&prod._final
group by segment;
quit;

data &prod._pd_coh_seg;
set &prod._pd_coh_seg;
abs_diff_pit=abs(dr_ct-pd_ct_pit);
abs_diff_ttc=abs(dr_ct-pd_ct_ttc);
run;

proc sort data =&prod._pd_coh_seg;
by ASOF_YYYYMM segment;
quit;

proc sql;
create table &prod._pd_mad_rov as   
select distinct
segment,
mean(rcpa_b2_pd_low),
mean(rcpa_b2_pd_high),
count(*) as Count,
sum(abs_diff_pit)/count(abs_diff_pit) as mad_pit,
sum(abs_diff_ttc)/count(abs_diff_ttc) as mad_ttc,
sum(order)/count(abs_diff_pit) as breaks
from &prod._pd_coh_seg
group by segment;
quit;

%mend;

%bktest(prod=auto);
%bktest(prod=cred);
%bktest(prod=manu);
%bktest(prod=spec);

%macro bktest(prod=);
PROC EXPORT DATA= &prod._pd_mad_rov
            OUTFILE= "/ecm_dr/retail/other/zkt0jqm/2014Q3/pd_mad_rov.xlsx" 
            DBMS=XLSX REPLACE;
     SHEET="&prod."; 
RUN;
%mend;

%bktest(prod=auto);
%bktest(prod=spec);
%bktest(prod=manu);
%bktest(prod=cred);

And the error message is:

ERROR: The MEAN summary function requires a numeric argument.
ERROR: The MEAN summary function requires a numeric argument.
ERROR: The following columns were not found in the contributing tables: rcpa_b2_pd_high, rcpa_b2_pd_low.

I don't understand where was the mistake. If I remove the mean statements on rcpa_b2_pd_high , rcpa_b2_pd_low , and group by segment, rcpa_b2_pd_high , rcpa_b2_pd_low, then this issue will be gone. However, I'm not sure that's the right methodology since these two are a rate percentage which shouldn't be in the group statement. Could anyone help me diagnose this issue? Your input is greatly appreciated.

SP

Answer 1

The issue is in the first SQL statement.In the first sql statement.

proc sql;

create table &prod._pd_coh_seg as   
select distinct
ASOF_YYYYMM,
segment,
mean(rcpa_b2_pd_low) as rcpa_b2_pd_low,
mean(rcpa_b2_pd_high) as rcpa_b2_pd_high,
sum(rcpa_weight) as Count,
sum(annual_default_occurrence*rcpa_weight)/sum(rcpa_weight) as dr_ct,
sum(dp_pit*rcpa_weight)/sum(rcpa_weight) as pd_ct_pit,
sum(RCPA_B2_PD_TTC_SEG_FCTR*rcpa_weight)/sum(rcpa_weight) as pd_ct_ttc
from backtest.&prod._final
group by segment;
quit;

this happens because when an aggregation is done the new aggregated column gets the random name assigned from sas