percentiles -not get all of them

[Ronein]

Hello

I have a data set with 35000 rows.

In this data set there are following columns:

"Customer ID"

"predicted" with value between 0 and 1 (probability of failure) 

"Failure" with value 1 or 0 (1 if customer fail in 12 months follow up and  0 if not)

I want to create a new data set with calculation of percentiles  1 until 100  ( 1,2,3,4,5,6,7,8,.......and so on till 100)

The problem:

In the new column "Percentile_"  there is jumping in the values and I get only the following percentiles

0
1
2
3
4
5
6
7
8
9
11
13
14
15
16
27
38
40
42
43
44
45
46
48
51
52
53
54
55
57
59
60
66
73
74
75
76
78
80
84
88
89
90
91
93
94
95
97
98
99

proc rank data=Have  groups=100 descending out=b;
var predicted;
ranks Percentile_;
run;

proc summary data=b nway;
var Failure ;
class Percentile_;
output out=d(drop=_type_ rename=(_freq_=n_Percentile)) 
sum=Bad_Percentile ;
run;

[Patrick]

I assume you've got these "jumps" because of ties. There are different options for Proc Rank how to treat ties.

https://go.documentation.sas.com/doc/en/pgmsascdc/9.4_3.5/proc/n0y7jshlj1odi6n1atnqzerl3uw5.htm 

[Ronein]

Sorry,I didnt see solution to my problem in this link

[PaigeMiller]

I want to create a new data set with calculation of percentiles 1 until 100 ( 1,2,3,4,5,6,7,8,.......and so on till 100)

Percentiles of what?

Do you understand that percentiles of 0/1 variables is a meaningless thing to compute?

[Ronein]

Thanks,

Percentiles of  column "predicted"  (Column predicted  get values between 0 and 1 and it is the probability of failure by model)

[Ronein]

Percentile of PD (probability of default)

[Ksharp]

As Patrick said try option ties=dense

proc rank data=Have groups=100 descending out=b TIES=DENSE ;

Or you could use PROC UNIVARIATE + PCTLPTS + PCTLPRE

[Ronein]

TIES=DENSE is not working well.
It classified to groups in wrong way.
May you please show the code of using PROC UNIVARIATE + PCTLPTS + PCTLPRE?

[PaigeMiller]

---

@Ronein wrote:

TIES=DENSE is not working well.

---

Show us what is wrong. Don't say something is wrong and provide no explanation.

It classified to groups in wrong way.

I think it classified the group in the correct way (it did what it was supposed to do), but that may not be what you want. And you need to explain what you want.