topic Obtaining cum. freqs for all possible outcomes in TABLES variable for PROC FREQ in SAS Procedures

Obtaining cum. freqs for all possible outcomes in TABLES variable for PROC FREQ

AvocadoRivalry — Fri, 03 Feb 2012 15:28:10 GMT

I am trying to build a dataset that shows the cumulative % of "bads" across 10 different buckets using PROC FREQ. I'm running into a problem for the buckets where there are zero "bads." SAS outputs what I've labeled below as "Table 1" where what I need is "Table 2." Basically, the difference that I'm trying to solve is to have SAS output that 80% - 90% bucked and output the same Cumulative Bad % from the bucket above it. Are there any options in PROC FREQ to do this? If not, what is the best alternative approach? Thanks!

TABLE 1:

Percentile	Bads	Cum. Bad %
0 - 10%	5	20.8%
10% - 20%	6	45.8%
20% - 30%	2	54.2%
30% - 40%	1	58.3%
40% - 50%	3	70.8%
50% - 60%	1	75.0%
60% - 70%	2	83.3%
70% - 80%	2	91.7%
90% - 100%	2	100.0%

TABLE 2:

Percentile	Bads	Cum. Bad %
0 - 10%	5	20.8%
10% - 20%	6	45.8%
20% - 30%	2	54.2%
30% - 40%	1	58.3%
40% - 50%	3	70.8%
50% - 60%	1	75.0%
60% - 70%	2	83.3%
70% - 80%	2	91.7%
80% - 90%	0	91.7%
90% - 100%	2	100.0%

Obtaining cum. freqs for all possible outcomes in TABLES variable for PROC FREQ

art297 — Fri, 03 Feb 2012 16:06:22 GMT

Take a look at the following SAS-L thread: http://listserv.uga.edu/cgi-bin/wa?A2=ind1107e&L=sas-l&D=0&P=3180

Re: Obtaining cum. freqs for all possible outcomes in TABLES variable for PROC FREQ

data_null__ — Fri, 03 Feb 2012 16:25:22 GMT

Another complete example with data. Uses CLASSDATA option.


proc rank data=sashelp.heart groups=10 out=smoke;
   var Smoking;
   run;
data class;
   if 0 then set smoke;
   do smoking=0 to 9;
      output;
      end;
   stop;
   run;
 
proc summary data=smoke nway classdata=class;
   class smoking;
   output out=summary;
   run;
 
proc freq;
   tables smoking;
   weight _freq_ / zeros;
   run;

Values of Smoking Were Replaced by Ranks

Cumulative Cumulative

Smoking Frequency Percent Frequency Percent

ƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒ

0 0 0.00 0 0.00

1 0 0.00 0 0.00

2 2501 48.35 2501 48.35

3 0 0.00 2501 48.35

4 113 2.18 2614 50.53

5 466 9.01 3080 59.54

6 576 11.13 3656 70.67

7 921 17.80 4577 88.48

8 125 2.42 4702 90.90

9 471 9.10 5173 100.00

Obtaining cum. freqs for all possible outcomes in TABLES variable for PROC FREQ

AvocadoRivalry — Fri, 03 Feb 2012 16:37:11 GMT

Easy enough. Using the preloadfmt / completetypes options in PROC Summary and then running that dataset through PROC Freq seems to get me exactly where I need. Thanks for the link!