How to calcuate median in proc report

[sas_student1]

Is it possible to calculate median in Proc report?

I have a dataset that has a client ID, years, service type and length of stay (los).

What I want to do is calculate using proc report (or maybe tabulate??) the median length of stay by service type for each year.

Is that possible or do I have to do this in proc means? or is there another way of doing this?

Asking because I am trying to use proc report to have columns that count the number of '1's under each year and then have another column that will calculate the median los by service type for each year. You will notice in my sample below that no two years have a '1' in the same row.

data have;
input ID $ cy2014 cy2015 cy2016 cy2017 cy2018 cy2019 service_type $ los;
datalines;
1	0	0	0	0	1	0	Outpatient	313
2	0	1	0	0	0	0	Hospital	14
3	0	0	0	0	0	1	Outpatient	59
4	0	0	0	0	1	0	Outpatient	104
5	0	0	0	0	0	1	Outpatient	57
6	0	0	0	0	0	1	Outpatient	34
7	0	0	0	0	1	0	ER	6
8	0	0	1	0	0	0	ER	4
9	0	0	1	0	0	0	HomeCare	332
10	1	0	0	0	0	0	Outpatient	25
11	0	0	0	0	1	0	Clinic	624
12	0	0	1	0	0	0	Outpatient	41
13	1	0	0	0	0	0	ER	4
14	0	1	0	0	0	0	Outpatient	95
15	0	0	0	1	0	0	ER	7
16	0	0	0	1	0	0	ER	29
17	1	0	0	0	0	0	Outpatient	258
18	0	0	0	1	0	0	Outpatient	36
19	0	0	0	0	0	1	HomeCare	38
20	0	0	0	0	1	0	Outpatient	362
21	0	0	0	0	1	0	Outpatient	107
22	1	0	0	0	0	0	Outpatient	252
23	0	0	0	0	1	0	Hospital	14
24	0	0	0	0	1	0	Outpatient	15
25	1	0	0	0	0	0	HomeCare	76
26	0	1	0	0	0	0	HomeCare	21
27	0	0	0	0	0	1	ER	3
28	0	1	0	0	0	0	Clinic	261
29	1	0	0	0	0	0	Outpatient	187
30	0	0	0	0	0	1	Outpatient	75
;

This is the code that I am using to count the '1's for each year by each service type. How can I add to this to a code to calculate the median los by each service type and each year?

proc report data=have ;
column service_type  cy2014 cy2015 cy2016 cy2017 cy2018 cy2019;
define service_type / group  width=50 'Program Type';
define cy2014 / sum '2014';
define cy2015 / sum '2015';
define cy2016 / sum '2016';
define cy2017 / sum '2017';
define cy2018 / sum '2018';
define cy2019 / sum '2019(YTD)';
run;

Thanks in Advance!

[PaigeMiller]

According to this, you can compute medians in PROC REPORT

https://documentation.sas.com/?cdcId=pgmsascdc&cdcVersion=9.4_3.4&docsetId=proc&docsetTarget=p1hagdk...

Just replace SUM with MEDIAN in your code.

[Cynthia_sas]

Hi:
I am not sure what your LOS variable is used for, but you don't list it on the report.

You say you want this:
"length of stay by service type for each year."
But you don't show length of stay for each year. I only see one value for LOS -- assuming that is length of stay.

To get the sum for each column is easy, you just need to add an RBREAK AFTER / SUMMARIZE; statement to your PROC REPORT code.

Since I don't understand how you are going to calculate Length of Stay, that is harder to make any constructive suggestions, since LOS is not on the report.
Cynthia

[sas_student1]

Hi,

Yes LOS (length of stay) is only showing by service type. But if you note no two years have a 1 indicator in the same row. so if there is a 1 in cy2014 you will note that none of the other years have a 1. Is it possible that in proc report a statement such as "if cy2014=1 then calculate median by service type and label the median as los for cy2014" and then  "if cy2015=1 then calculate median by service type and label the median as los for cy2015" and so far?

I am trying to get to the bottom table.

LOS is already calculate I just want to try to calculate the median LOS by each service type by the year. So something like the below as the result.

Program Type	2014	Median LOS 2014	2015	Median LOS2015	2016	Median LOS 2016	2017	Median LOS 2017	2018	Median LOS 2018	2019	Median LOS 2019
Clinic	0		1	261	0		0		1	624	0
ER	1	4	0		1	4	2	18	1	6	1	3
HomeCare	1	76	1	21	1	332	0		0		1	38
Hospital	0		1	14	0		0		1	14	0
Outpatie	4	219.5	1	95	1	41	1	36	5	107	4	58

[Cynthia_sas · Accepted solution]

Hi:

  PROC REPORT could do it, but it will be far easier to do it in the DATA step where you read the data using ARRAY processing. I think you want something like this:

Which I accomplished with this:

data fakedata;
  length service_type $10;
  input ID $ cy2014 cy2015 cy2016 cy2017 cy2018 cy2019 service_type $ los;
  
  ** use array processing to create length of stay variables;
  ** for each year. IF "CYxxxx" variable is 1 then STAYxxxx variable;
  ** is assigned the LOS value for that observation;
  array arrcy cy2014-cy2019;
  array arrlos stay2014-stay2019;
  do i = 1 to dim(arrcy);
     if arrcy(i) = 1 then arrlos(i) = los;
	 else if arrcy(i) = 0 then arrlos(i) = 0;
  end;
datalines;
1	0	0	0	0	1	0	Outpatient	313
2	0	1	0	0	0	0	Hospital	14
3	0	0	0	0	0	1	Outpatient	59
4	0	0	0	0	1	0	Outpatient	104
5	0	0	0	0	0	1	Outpatient	57
6	0	0	0	0	0	1	Outpatient	34
7	0	0	0	0	1	0	ER	6
8	0	0	1	0	0	0	ER	4
9	0	0	1	0	0	0	HomeCare	332
10	1	0	0	0	0	0	Outpatient	25
11	0	0	0	0	1	0	Clinic	624
12	0	0	1	0	0	0	Outpatient	41
13	1	0	0	0	0	0	ER	4
14	0	1	0	0	0	0	Outpatient	95
15	0	0	0	1	0	0	ER	7
16	0	0	0	1	0	0	ER	29
17	1	0	0	0	0	0	Outpatient	258
18	0	0	0	1	0	0	Outpatient	36
19	0	0	0	0	0	1	HomeCare	38
20	0	0	0	0	1	0	Outpatient	362
21	0	0	0	0	1	0	Outpatient	107
22	1	0	0	0	0	0	Outpatient	252
23	0	0	0	0	1	0	Hospital	14
24	0	0	0	0	1	0	Outpatient	15
25	1	0	0	0	0	0	HomeCare	76
26	0	1	0	0	0	0	HomeCare	21
27	0	0	0	0	0	1	ER	3
28	0	1	0	0	0	0	Clinic	261
29	1	0	0	0	0	0	Outpatient	187
30	0	0	0	0	0	1	Outpatient	75
;
run;
  
title; footnote;
proc report data=fakedata ;
column service_type  
       ('2014' cy2014 stay2014) ('2015' cy2015 stay2015) 
       ('2016' cy2016 stay2016) ('2017' cy2017 stay2017) 
       ('2018' cy2018 stay2018) ('2019(YTD)' cy2019 stay2019) ;
define service_type / group  'Program Type';
define cy2014 / sum "CY";
define stay2014 / sum "LOS";
define cy2015 / sum "CY";
define stay2015 / sum "LOS";
define cy2016 / sum "CY";
define stay2016 / sum "LOS";
define cy2017 / sum  "CY";
define stay2017 / sum  "LOS";
define cy2018 / sum "CY";
define stay2018 / sum "LOS";
define cy2019 / sum "CY";
define stay2019 / sum "LOS";
 
rbreak after / summarize;
run;

Hope this helps,

cynthia

[sas_student1]

Ah! I see what you did there! Thank you!!!