topic Re: How do I create a line graph showing proportion, by age, over time using PROC SGPLOT? in SAS Programming

How do I create a line graph showing proportion, by age, over time using PROC SGPLOT?

_maldini_ — Thu, 31 Mar 2022 18:51:45 GMT

I'd like to create a line graph showing the proportion of respondents answering "Yes" to past month tobacco use, by age category, over time. Basically I am trying to create a graph like this one:

These instructions are helpful, but I'm not sure how to manipulate the data before running the PROC SGPLOT.

When I run this syntax using the data as is...


PROC SGPLOT DATA = input;
   SERIES X = year Y = past_month_use / group=age_cat;
   SERIES X = year Y = past_month_use / group=age_cat;
   SERIES X = year Y = past_month_use / group=age_cat;
RUN;

I get this:

How do I need to manipulate these data so that I can produce the desired outcome?

Thank you.

Sample data below:
4 vars:
id (1-20 respondents)
age_cat (1,2,3)
past_month_use (0=no, 1=yes)
year (2018-2020)

id age_cat past_month_use year

1 1 1 2018

1 1 1 2019

1 1 1 2020

1 2 0 2021

2 2 0 2018

2 2 0 2019

2 2 1 2020

2 2 1 2021

3 3 1 2018

3 3 1 2019

3 3 1 2020

3 3 1 2021

4 2 1 2018

4 3 1 2019

4 3 1 2020

4 3 1 2021

5 1 0 2018

5 1 0 2019

5 1 0 2020

5 1 0 2021

6 3 0 2018

6 3 1 2019

6 3 0 2020

6 3 1 2021

7 2 1 2018

7 2 1 2019

7 2 1 2020

7 2 1 2021

8 2 0 2018

8 2 1 2019

8 3 0 2020

8 3 1 2021

9 3 1 2018

9 3 1 2019

9 3 1 2020

9 3 1 2021

10 1 1 2018

10 1 0 2019

10 1 0 2020

10 1 0 2021

11 2 0 2018

11 2 0 2019

11 2 1 2020

11 2 1 2021

12 1 1 2018

12 1 0 2019

12 1 1 2020

12 2 1 2021

13 3 1 2018

13 3 1 2019

13 3 1 2020

13 3 1 2021

14 2 1 2018

14 2 1 2019

14 2 0 2020

14 3 0 2021

15 1 0 2018

15 1 0 2019

15 1 1 2020

15 1 1 2021

16 2 1 2018

16 2 0 2019

16 2 0 2020

16 2 0 2021

17 2 0 2018

17 3 1 2019

17 3 1 2020

17 3 1 2021

18 2 1 2018

18 2 1 2019

18 2 1 2020

18 2 1 2021

19 3 1 2018

19 3 1 2019

19 3 1 2020

19 3 1 2021

20 1 1 2018

20 2 1 2019

20 2 1 2020

20 2 1 2021

Re: How do I create a line graph showing proportion, by age, over time using PROC SGPLOT?

yabwon — Thu, 31 Mar 2022 20:16:45 GMT

Do you mean something like this:

data have;
input id age_cat past_month_use year;
cards;
1 1 1 2018
1 1 1 2019
1 1 1 2020
1 2 0 2021
2 2 0 2018
2 2 0 2019
2 2 1 2020
2 2 1 2021
3 3 1 2018
3 3 1 2019
3 3 1 2020
3 3 1 2021
4 2 1 2018
4 3 1 2019
4 3 1 2020
4 3 1 2021
5 1 0 2018
5 1 0 2019
5 1 0 2020
5 1 0 2021
6 3 0 2018
6 3 1 2019
6 3 0 2020
6 3 1 2021
7 2 1 2018
7 2 1 2019
7 2 1 2020
7 2 1 2021
8 2 0 2018
8 2 1 2019
8 3 0 2020
8 3 1 2021
9 3 1 2018
9 3 1 2019
9 3 1 2020
9 3 1 2021
10 1 1 2018
10 1 0 2019
10 1 0 2020
10 1 0 2021
11 2 0 2018
11 2 0 2019
11 2 1 2020
11 2 1 2021
12 1 1 2018
12 1 0 2019
12 1 1 2020
12 2 1 2021
13 3 1 2018
13 3 1 2019
13 3 1 2020
13 3 1 2021
14 2 1 2018
14 2 1 2019
14 2 0 2020
14 3 0 2021
15 1 0 2018
15 1 0 2019
15 1 1 2020
15 1 1 2021
16 2 1 2018
16 2 0 2019
16 2 0 2020
16 2 0 2021
17 2 0 2018
17 3 1 2019
17 3 1 2020
17 3 1 2021
18 2 1 2018
18 2 1 2019
18 2 1 2020
18 2 1 2021
19 3 1 2018
19 3 1 2019
19 3 1 2020
19 3 1 2021
20 1 1 2018
20 2 1 2019
20 2 1 2020
20 2 1 2021
;
run;

proc sort data = have;
  by year age_cat past_month_use id;
run;

data want;
  set have;
  by year age_cat;
  if first.age_cat then
    do;
      d = 0;
      n = 0;
    end;

  d + past_month_use;
  n + 1;

  if last.age_cat then
    do;
      prop = divide(d,n);
      if prop then output;
      format prop percent10.2;
    end;
run;

proc sgplot data = want;
  series x = year y = prop / group=age_cat;
  xaxis integer;
run;

Bart

Re: How do I create a line graph showing proportion, by age, over time using PROC SGPLOT?

ballardw — Thu, 31 Mar 2022 20:23:51 GMT

You need to summarize the data to create the proportion.

Using a 1/0 coded value when you use the mean you get the percentage (proportion) of 1 values. I used proc summary, other methods are possible, to create data set with the summary and the other values.

The Xaxis and Yaxis statements to make the X values "nicer" and to force 0 into the graph for the Y value axis.

The format is very optional but sets a nice appearance. Labels on your variables would get nicer axis labels.

data have;
  input id age_cat past_month_use year;
datalines;
1 1 1 2018
1 1 1 2019
1 1 1 2020
1 2 0 2021
2 2 0 2018
2 2 0 2019
2 2 1 2020
2 2 1 2021
3 3 1 2018
3 3 1 2019
3 3 1 2020
3 3 1 2021
4 2 1 2018
4 3 1 2019
4 3 1 2020
4 3 1 2021
5 1 0 2018
5 1 0 2019
5 1 0 2020
5 1 0 2021
6 3 0 2018
6 3 1 2019
6 3 0 2020
6 3 1 2021
7 2 1 2018
7 2 1 2019
7 2 1 2020
7 2 1 2021
8 2 0 2018
8 2 1 2019
8 3 0 2020
8 3 1 2021
9 3 1 2018
9 3 1 2019
9 3 1 2020
9 3 1 2021
10 1 1 2018
10 1 0 2019
10 1 0 2020
10 1 0 2021
11 2 0 2018
11 2 0 2019
11 2 1 2020
11 2 1 2021
12 1 1 2018
12 1 0 2019
12 1 1 2020
12 2 1 2021
13 3 1 2018
13 3 1 2019
13 3 1 2020
13 3 1 2021
14 2 1 2018
14 2 1 2019
14 2 0 2020
14 3 0 2021
15 1 0 2018
15 1 0 2019
15 1 1 2020
15 1 1 2021
16 2 1 2018
16 2 0 2019
16 2 0 2020
16 2 0 2021
17 2 0 2018
17 3 1 2019
17 3 1 2020
17 3 1 2021
18 2 1 2018
18 2 1 2019
18 2 1 2020
18 2 1 2021
19 3 1 2018
19 3 1 2019
19 3 1 2020
19 3 1 2021
20 1 1 2018
20 2 1 2019
20 2 1 2020
20 2 1 2021
;

proc summary data=have nway;
   class age_cat year;
   var past_month_use ;
   output out=summary mean=;
run;

proc sgplot data=summary;
   series x=year y=past_month_use/ group=age_cat;
   xaxis values=(2018 to 2021 by 1) ;
   yaxis values=(0 to 1 by .1);
   format past_month_use percent6.;
run;

Re: How do I create a line graph showing proportion, by age, over time using PROC SGPLOT?

_maldini_ — Thu, 31 Mar 2022 21:46:27 GMT

Thank you!

Can you accomplish the same thing using PROC FREQ instead of PROC SUMMARY?

Re: How do I create a line graph showing proportion, by age, over time using PROC SGPLOT?

ballardw — Thu, 31 Mar 2022 22:20:33 GMT

@_maldini_ wrote:

Thank you!

Can you accomplish the same thing using PROC FREQ instead of PROC SUMMARY?

Suggestion is always "try and see".

The main thing with proc freq is it COUNTS all formatted values. So you would end up with counts/percents of 0 as well as 1. Also Proc Freq shifts percentages calculated by *100 so the Y axis values change.

proc freq data=have noprint;
  tables age_cat*year*past_month_use / outpct out=freqout;

run;

proc sgplot data=freqout;
   where past_month_use=1;
   series x=year y=pct_row/ group=age_cat;
   xaxis values=(2018 to 2021 by 1) ;
   yaxis values=(0 to 100 by 10);
run;