Date: Wed, 31 Jan 2007 12:57:44 -0500
Reply-To: "Howard Schreier <hs AT dc-sug DOT org>" <nospam@HOWLES.COM>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: "Howard Schreier <hs AT dc-sug DOT org>" <nospam@HOWLES.COM>
Subject: Re: calculating percentages in proc report
Content-Type: text/plain; charset=ISO-8859-1
On Sun, 28 Jan 2007 16:42:25 -0500, Janet Carlson <janet.carlson@NDSU.EDU>
wrote:
>hi proc report experts
>I am new to proc report and need to put the results of a 500 variable
>survey into proc report. The survey is in sections so I am approaching it
>by trying to do this in sections. Example satisf1 – satisf15, var1 –
>var10, x1 – x12
>The format of the report would be:
>Var1 Male female total
> 25 20 30
> 15 10 15
> 20 30 20
> 10 25 10
> 30 15 25
> 100 100 100
>
>Same for var2 – var10
>Then x1 – x12
>Then satisf1 – satisf15 ect.
>
>I am able to get the total column to sum to 100 but have been unable to
>get the male and female columns to sum to 100. everything I try I get the
>percentages for male and female to sum to 100 for the entire section not
>the individual variable.
>Following is a small data set and my attempt at proc report. I am
>changing the sections of variables into a univariate structure. I need
>to know how to get the male and female totals to sum to 100 by each
>variable. Also any suggestions are welcome. Should I approach this
>differently? Like I said this is a first attempt at working with proc
>report.
>Thank you.
[snip]
I've made a bunch of non-critical changes to the original code to eliminate
some distractions, make things less roundabout, and make the output more
compact.
data yfcy06;
input obs sex06a $ intact1-intact4;
cards;
1 F 2 1 2 2
2 F 3 2 3 3
3 F 4 3 3 4
4 F 5 1 2 1
5 F 5 5 4 4
6 F 3 5 2 2
7 F 2 2 2 2
8 F 2 3 4 5
9 F 1 1 1 1
10 F 1 2 2 5
11 M 3 1 2 1
12 M 2 3 1 4
13 M 3 2 1 .
14 M 2 1 2 2
15 M 1 3 2 2
16 M 4 2 2 2
17 M 5 5 3 3
18 M 5 5 4 5
19 M 4 2 3 4
20 M 5 5 2 2
;
proc transpose data=yfcy06
out=aa(rename = (_name_=qall col1=intact) );
by obs sex06a;
run;
data aaa; set aa;
intacta=intact;
label qall = 'Who/Where';
run;
proc format;
value $sex
F='Male'
M='Female';
value intact
1='Never'
2='1-2 / term'
3='1-2 / month'
4='1 / week'
5='2-3 / week'
6='Daily';
value $intactq
'intact1' = 'Faculty in'
'intact2' = 'Faculty out'
'intact3' = 'Advisors'
'intact4' = 'Other';
run;
proc report nowindows ;
column qall intact sex06a, (pctn) intacta n pctntot;
define qall / group order=data format= $intactq.;
define intact / group order=internal format=intact.;
define sex06a / across format=$sex.;
define intacta / analysis n noprint;
define pctntot / computed format=percent8.1;
define pctn / format=percent8.1;
compute pctntot;
pctntot = intacta.n / pctnn;
endcomp;
compute before qall;
pctnn = intacta.n; endcomp;
break after qall / skip summarize;
run;
If I understand correctly, Janet wants the denominators for the PCTN stats
to be at the QALL group level. I see nothing in the code to make that happen
and I'm not a REPORT export, so I have no suggestion.
Meanwhile, the output looks like this:
sex06a
Who Female Male
Where intact pctn pctn n pctntot
Faculty in Never 2.6% 5.0% 3 15.0%
1-2 / term 5.1% 7.5% 5 25.0%
1-2 / month 5.1% 5.0% 4 20.0%
1 / week 5.1% 2.5% 3 15.0%
2-3 / week 7.7% 5.0% 5 25.0%
Faculty in 25.6% 25.0% 20 100.0%
Faculty out Never 5.1% 7.5% 5 25.0%
1-2 / term 7.7% 7.5% 6 30.0%
1-2 / month 5.1% 5.0% 4 20.0%
2-3 / week 7.7% 5.0% 5 25.0%
Faculty out 25.6% 25.0% 20 100.0%
Advisors Never 5.1% 2.5% 3 15.0%
1-2 / term 12.8% 12.5% 10 50.0%
1-2 / month 5.1% 5.0% 4 20.0%
1 / week 2.6% 5.0% 3 15.0%
Advisors 25.6% 25.0% 20 100.0%
Other Never 2.6% 5.0% 3 15.8%
1-2 / term 10.3% 7.5% 7 36.8%
1-2 / month 2.6% 2.5% 2 10.5%
1 / week 5.1% 5.0% 4 21.1%
2-3 / week 2.6% 5.0% 3 15.8%
Other 23.1% 25.0% 19 100.0%
The 7.5% in the first row should be 30%, etc.
