topic Percentage accuracy in proc freq table in SAS Procedures

Percentage accuracy in proc freq table

Sathish_jammy — Wed, 07 Aug 2019 06:48:52 GMT

Dear Experts,

I couldn't get the 100% in freq of the age column. Kindly go through the result below for more info.

Age_cat	Frequency	Percent
18-25	120	13.33
26-35	180	20
36-45	180	20
46-55	180	20
56-65	120	13.33
>65	120	13.33

If I sum up the percentage it returns 99.99. There is no missing/ value 0 in the age column. How could I solve it, Please suggest some ideas?

Thanks in advance!

Re: Percentage accuracy in proc freq table

Kurt_Bremser — Wed, 07 Aug 2019 06:58:55 GMT

That's the math of limited digits. Add one decimal digit for the display (percent8.3), then your manual addition will result in 99.999, and when you round that to 0.01, you get 100.

Re: Percentage accuracy in proc freq table

Reeza — Wed, 07 Aug 2019 07:01:20 GMT

Add another decimal place to the output. It’s due to rounding into thirds which doesn’t work evenly ever.

@Sathish_jammy wrote:

Dear Experts,

I couldn't get the 100% in freq of the age column. Kindly go through the result below for more info.

Age_cat Frequency Percent

18-25 120 13.33

26-35 180 20

36-45 180 20

46-55 180 20

56-65 120 13.33

>65 120 13.33

If I sum up the percentage it returns 99.99. There is no missing/ value 0 in the age column. How could I solve it, Please suggest some ideas?

Thanks in advance!

Re: Percentage accuracy in proc freq table

ballardw — Wed, 07 Aug 2019 15:21:58 GMT

It may help to share data and the actual code used. Normally Proc Freq will show two decimals for every percentage.

Example:

proc freq data=sashelp.class;
   tables age;
run;

yields

                                Cumulative    Cumulative
Age    Frequency     Percent     Frequency      Percent
ƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒ
 11           2       10.53             2        10.53
 12           5       26.32             7        36.84
 13           3       15.79            10        52.63
 14           4       21.05            14        73.68
 15           4       21.05            18        94.74
 16           1        5.26            19       100.00

So, since you are not showing any decimals for the 20 percent values you have done something other than default. That might be the cause. With a data set that would replicate your behavior:

data junk;
  input cat freq;
datalines;
1 120
2 180
3 180
4 180
5 120
6 120
;
run;

proc freq data=junk;
   tables cat/ nocum out=work.summary;
   weight freq;
run;

we get

cat    Frequency     Percent
ƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒ
  1         120       13.33
  2         180       20.00
  3         180       20.00
  4         180       20.00
  5         120       13.33
  6         120       13.33

and if you examine the created output set work.summary you will find there are more decimal places as each of your categories with 120 frequency will have repeating decimals of .33333333 etc.

So summing those will approach 100.

Or a different procedure:

proc tabulate data=junk;
  class cat;
  freq freq;
  tables cat All='Total',
         n pctn;
run;

But where or however you "sum" the percentage column manually you need to actually know the actual values involved.