topic Re: GROTTA BAR in SAS Programming

GROTTA BAR

desireatem — Mon, 16 Aug 2021 16:13:06 GMT

I have the following data and wish to produce the graph below where type is 0=no and 1=yes.

Please any help is more than welcome.

Type mrs (90 days outcome)

0 5

1 1

1 2

1 0

0 3

1 6

0 4

Re: GROTTA BAR

PaigeMiller — Mon, 16 Aug 2021 16:17:22 GMT

PROC SGPLOT with the HBAR statement.

Re: GROTTA BAR

desireatem — Mon, 16 Aug 2021 16:31:49 GMT

The Hbar statement won't solve it since it will simple have mrs values on the y -axis rather than type ("no" "yes")

Re: GROTTA BAR

PaigeMiller — Mon, 16 Aug 2021 16:35:11 GMT

Use HBAR TYPE

Re: GROTTA BAR

desireatem — Mon, 16 Aug 2021 16:45:00 GMT

Thank you, this is extremely helpful. How do I put the number into the different shaded regions as in the plot.

Re: GROTTA BAR

PaigeMiller — Mon, 16 Aug 2021 16:49:58 GMT

I'm still not 100% sure what you are doing, as the numbers in your table don't seem to match the numbers in the plot.

There are a number of options in the HBAR documentation segment named "Label Options", probably one of them does what you want.

Re: GROTTA BAR

desireatem — Mon, 16 Aug 2021 16:58:19 GMT

That is just part of the data. In the grotta bar graph that I pasted for type "no" we have number 88, 7, 41... How do I have similar plot with different n.??

Re: GROTTA BAR

PaigeMiller — Mon, 16 Aug 2021 17:01:32 GMT

Explain what number you want in the bar chart. How is this number derived/computed from the data you show?

Re: GROTTA BAR

desireatem — Mon, 16 Aug 2021 17:17:51 GMT

For instant for type =no, if 88 people had mrs=1, the number 88 will show within the box in the plot, if 7 had mrs=2, the number 2 will appear in corresponding color within the box. That is how the figure above is. You see the sample size within each color, corresponding to the number of people with that mrs value

Re: GROTTA BAR

Reeza — Mon, 16 Aug 2021 22:48:25 GMT

Both of your bars there are the exact same length which implies you're showing them as a stacked percentage but labelling them with counts. Is that how you want it to show or do you want the bar lengths to be different. If you want it as shown, I suspect you may need to pre-process the data to have it work properly because the labels do not match the axis so it's not using the pre-calculated statistics. And since a count won't align with a stacked 100% chart (they would if not 100%) then I suspect you need to pre-calculate things.

Re: GROTTA BAR

Reeza — Mon, 16 Aug 2021 23:35:39 GMT

This should answer your question. There are two options below, not really sure which one you're after. If they don't answer it exactly they should be pretty close.
If they do not work please be very very specific about how.

data random_data;

call streaminit(99);

do RECORD_NUM = 1 to 100;
	TYPE = rand("bernoulli", 0.5);
	mRS = rand('integer', 0, 6);
	output;
end;
run;

proc format;
value mRS_fmt 
0 -1 = "mRS 0 - 1"
2 = "mRS 2"
3-5 = "mRS 5"
6 = "mRS 6";
run;

proc freq data=random_data;
table Type*MRS / out=summary_data outpct;
format mRS mRS_fmt.;
run;

proc sgplot data=summary_data;
hbar type / group = mRS response = pct_row stat=pct;
*to get counts labelled;
hbar type / group = mRS response = count stat=sum seglabel x2axis;
x2axis display=none;
run;

proc sgplot data=random_data;
hbar type / group = mRS stat=pct;
hbar type / group = mRS stat=freq seglabel x2axis;
x2axis display=none;
format mRS mRS_fmt.;
run;

@desireatem wrote:

I have the following data and wish to produce the graph below where type is 0=no and 1=yes.

Please any help is more than welcome.

Type     mrs (90 days outcome)

0             5

1          1

1            2

1          0

0 3

1 6

0       4

Re: GROTTA BAR

desireatem — Tue, 17 Aug 2021 17:21:46 GMT

This is great! Thanks.

Re: GROTTA BAR

Reeza — Tue, 17 Aug 2021 17:47:27 GMT

lol, which answer was correct, the first SGPLOT or the second? Your question was very unclear so this was a lucky guess.

Re: GROTTA BAR

desireatem — Tue, 17 Aug 2021 18:35:22 GMT

This is the correct answer:

data random_data;

call streaminit(99);

do RECORD_NUM = 1 to 100;
	TYPE = rand("bernoulli", 0.5);
	mRS = rand('integer', 0, 6);
	output;
end;
run;

proc format;
value mRS_fmt 
0 -1 = "mRS 0 - 1"
2 = "mRS 2"
3-5 = "mRS 5"
6 = "mRS 6";
run;

proc freq data=random_data;
table Type*MRS / out=summary_data outpct;
format mRS mRS_fmt.;
run;

proc sgplot data=summary_data;
hbar type / group = mRS response = pct_row stat=pct;
*to get counts labelled;
hbar type / group = mRS response = count stat=sum seglabel x2axis;
x2axis display=none;
run;

proc sgplot data=random_data;
hbar type / group = mRS stat=pct;
hbar type / group = mRS stat=freq seglabel x2axis;
x2axis display=none;
format mRS mRS_fmt.;
run;

Re: GROTTA BAR

Reeza — Tue, 17 Aug 2021 18:49:12 GMT

There are two sgplots in that code that generate different graphs, so which one is actually right from what you wanted initially.....

Re: GROTTA BAR

desireatem — Wed, 18 Aug 2021 13:52:53 GMT

Both can be used but I like the second plots. So they are both great solutions.

Re: GROTTA BAR

desireatem — Wed, 18 Aug 2021 16:28:40 GMT

Hello Reeza. I marked this as correct but for some reason the colors in the grotta bar does not match the key below. For home and IRF both should have same color for 0, 1, 2...., 6. Also how do I change the name form mrs_discharge in the key to "mrs at discharge"

Re: GROTTA BAR

Reeza — Wed, 18 Aug 2021 16:59:45 GMT

Show your full code with sample data. Obviously that isn't my code otherwise the categories would be as shown in the FORMAT statement.

Re: GROTTA BAR

desireatem — Wed, 18 Aug 2021 16:59:54 GMT

proc format;
value mRS_fmt
0 = "mRS 0"
1 = "mRS 1"
2 = "mRS 2"
3 = "mRS 3"
4 = "mRS 4"
5 = "mRS 5"
6 = "mRS 6";
run;

proc format;
value disposition_fmt
0 = "Home"
3 = "IRF";
run;

proc freq data=IRF_sub;
table disposition_1*mrs_discharge/ out=summary_data outpct;
format mrs_discharge mRS_fmt.;
run;

title "Discharge mRS by Disposition";
proc sgplot data=summary_data;
hbar disposition_1 / group = mrs_discharge response = pct_row stat=pct;
*to get counts labelled;
hbar disposition_1 / group = mrs_discharge response = count stat=sum seglabel x2axis;
x2axis display=none;
format disposition_1 disposition_fmt.;
run;

Re: GROTTA BAR

Reeza — Wed, 18 Aug 2021 17:06:11 GMT

Yeah, doesn't seem like it's quite right. Can you please clarify the requirements based on the sample data I provided. Exactly what numbers do you expect to see in the bars based on that random data.