Display difference between means using VBOX

[MRM95]

Hello SAS users,

I have the following data, for which I am trying to build a specific boxplot.

I am nearly there, I just need to be able to display the difference between the connected means in each panel. Does anyone know how to do it? Many thanks!

Data:

data kccq_data;
length Treatment $ 15;
input ID $ KCCQ_var Treatment $ Timepoint;
datalines;
1 15 Implant 1
2 24 Implant 1
3 34 Implant 1
4 44 Implant 1
5 54 Implant 1
6 65 Implant 1
7 73 Implant 1
8 83 Implant 1
9 95.5 Implant 1
10 20 Implant 7
11 31 Implant 7
12 45.5 Implant 7
13 55.5 Implant 7
14 51 Implant 7
15 65.5 Implant 7
16 82 Implant 7
17 82 Implant 7
18 72 Implant 7
19 60 Implant 7
20 20 Implant 7
1 10 Control 1
2 22 Control 1
3 30 Control 1
4 42 Control 1
5 50 Control 1
6 80 Control 1
7 71 Control 1
8 60 Control 1
9 90 Control 1
10 72 Control 1
11 51 Control 7
12 30 Control 7
13 50 Control 7
14 70 Control 7
15 51 Control 7
16 60 Control 7
17 20 Control 7
18 11 Control 7
19 20 Control 7
20 30 Control 7
;
run;

Code to produce boxplot so far:

PROC SGPANEL DATA = kccq_data;
PANELBY Treatment / columns = 2 novarname;
VBOX KCCQ_var / category = Timepoint GROUP= Treatment connect=mean;
keylegend / title = "Treatment arm";
ROWAXIS label='KCCQ Overall Summary Score';
RUN;

[Ksharp]

/*
You need to calculated it by hand.
*/

data kccq_data;
length Treatment $ 15;
infile cards truncover;
input ID $ KCCQ_var Treatment $ Timepoint mean_diff;
label mean_diff='diff between means is ';
datalines;
1 15 Implant 1  100
2 24 Implant 1
3 34 Implant 1
4 44 Implant 1
5 54 Implant 1
6 65 Implant 1
7 73 Implant 1
8 83 Implant 1
9 95.5 Implant 1
10 20 Implant 7
11 31 Implant 7
12 45.5 Implant 7
13 55.5 Implant 7
14 51 Implant 7
15 65.5 Implant 7
16 82 Implant 7
17 82 Implant 7
18 72 Implant 7
19 60 Implant 7
20 20 Implant 7
1 10 Control 1    2000
2 22 Control 1
3 30 Control 1
4 42 Control 1
5 50 Control 1
6 80 Control 1
7 71 Control 1
8 60 Control 1
9 90 Control 1
10 72 Control 1
11 51 Control 7
12 30 Control 7
13 50 Control 7
14 70 Control 7
15 51 Control 7
16 60 Control 7
17 20 Control 7
18 11 Control 7
19 20 Control 7
20 30 Control 7
;
run;


PROC SGPANEL DATA = kccq_data;
PANELBY Treatment / columns = 2 novarname;
VBOX KCCQ_var / category = Timepoint GROUP= Treatment connect=mean;
keylegend / title = "Treatment arm";
ROWAXIS label='KCCQ Overall Summary Score';
inset mean_diff/position=ne;
RUN;

[LeliaM · Accepted solution]

You can use an INSET statement to display the difference value in the graph. Below is code that shows you how to calculate the difference.

data kccq_data;

length Treatment $ 15;

input ID $ KCCQ_var Treatment $ Timepoint;

datalines;

1 15 Implant 1

2 24 Implant 1

3 34 Implant 1

4 44 Implant 1

5 54 Implant 1

6 65 Implant 1

7 73 Implant 1

8 83 Implant 1

9 95.5 Implant 1

10 20 Implant 7

11 31 Implant 7

12 45.5 Implant 7

13 55.5 Implant 7

14 51 Implant 7

15 65.5 Implant 7

16 82 Implant 7

17 82 Implant 7

18 72 Implant 7

19 60 Implant 7

20 20 Implant 7

1 10 Control 1

2 22 Control 1

3 30 Control 1

4 42 Control 1

5 50 Control 1

6 80 Control 1

7 71 Control 1

8 60 Control 1

9 90 Control 1

10 72 Control 1

11 51 Control 7

12 30 Control 7

13 50 Control 7

14 70 Control 7

15 51 Control 7

16 60 Control 7

17 20 Control 7

18 11 Control 7

19 20 Control 7

20 30 Control 7

;

run;

proc sort data=kccq_data out=kccq_data;

 by treatment timepoint;

run;

 /* calculate mean for each treatment & timepoint */

proc means data=kccq_data;

 var kccq_var;

 by treatment timepoint;

 output out=mean mean=mean;

run;

proc sort data=mean out=result;

  by treatment;

run;

 /* Calculate Difference Between Mean Values For Each Treatment & Timepoint */

/* This code assumes there are two treatments. You can modify this code to make it more dynamic */

 data new (keep=diff treatment timepoint) ;

 label diff='Diff between means is ';

 format diff 5.2;

 set result;

   lag1=lag(mean);

   lag2=lag2(mean);

 if _n_=2 then do;

   diff=lag1-mean;

   output;

 end;

 if _n_=4 then do;

   diff=lag1-mean;

 output;

 end;

 run;

data all;

  merge new kccq_data;

  by treatment  ;

run;

proc sgpanel data = all;

  panelby treatment / columns = 2 novarname;

  vbox kccq_var / category = timepoint group= treatment connect=mean;

  keylegend / title = "Treatment Arm";

  rowaxis label='KCCQ Overall Summary Score';

  inset  diff/position=ne;

run;