topic Re: Quantile in SAS Enterprise Guide

Quantile

Stepik — Thu, 06 Jul 2023 11:47:10 GMT

Team, I tri to use the formula for calculating:

G = QUANTILE('NORMAL', pd). pd has different values, also of 1.

Unfortunately, the log gives me the next error

Does this mean that the function can not provide inverse cumulative distribution function of a real number?

Stepik

Re: Quantile

Stepik — Thu, 06 Jul 2023 11:47:39 GMT

of a real number = 1

Re: Quantile

Ksharp — Thu, 06 Jul 2023 12:21:31 GMT

Your argument of QUANTILE() is not right .

since it is a P value, it should be (0,1).

data have;
input pd;
cards;
.1
.2
.3
.4
;


data _null_;
set have;
G = QUANTILE('NORMAL', pd);
put G= pd=;
run;

38   ;
39
40
41   data _null_;
42   set have;
43   G = QUANTILE('NORMAL', pd);
44   put G= pd=;
45   run;

G=-1.281551566 pd=0.1
G=-0.841621234 pd=0.2
G=-0.524400513 pd=0.3
G=-0.253347103 pd=0.4
NOTE: 从数据集 WORK.HAVE. 读取了 4 个观测
NOTE: “DATA 语句”所用时间（总处理时间）:
      实际时间          0.01 秒
      CPU 时间          0.00 秒

Re: Quantile

Stepik — Thu, 06 Jul 2023 13:24:12 GMT

Solved taking pd=1 separately

Re: Quantile

Ksharp — Thu, 06 Jul 2023 12:38:20 GMT

Sorry. I have to leave now. I will take a look tomorrow.

But maybe @Rick_SAS could give you a hand. He is good at this.

Re: Quantile

Ksharp — Thu, 06 Jul 2023 12:40:20 GMT

And also better post sas dataset " R.Cov" . As I said you can't have pd_ap=1 in it.

Re: Quantile

FreelanceReinh — Thu, 06 Jul 2023 13:16:13 GMT

Hello @Stepik,

I think you just need to handle the cases pd=0 and pd=1 separately in the calculation of G and N:

if pd in (0 1) then N=pd;
else do;
  G = QUANTILE('NORMAL', pd); 
  N = CDF('NORMAL', G/SQRT(1-R)+SQRT(R/(1-R))*QUANTILE('NORMAL',0.999));
end;

Alternatively, since N=pd implies RWA1=0, you can also put the entire block of your assignment statements into the ELSE branch:

if pd in (0 1) then RWA1=0;
else do;
  ... your code ...
end;

(before the RETURN statement).

Re: Quantile

Stepik — Thu, 06 Jul 2023 13:17:06 GMT

I did it separately indeed for pd=1. thanks