Forest plot

[Abimal_Zippi]

Can you please advise how to clip a wide confidence interval in a forest plot? I am creating a forest plot and one of the variables had a wide 95% CL and would like to cut off the upper limit and show it by an arrow or by a continous indicator > etc...

Thanks

[ballardw]

Data please, in the form of a data step.

And the current code you are using to create your plot.

There are enough different ways to create any given plot that we need details to address your specific problem. Without data we can't test if the code does what is needed.

[Abimal_Zippi]

Thanks, I am trying to replicate the following code from Paper Template (sas.com)

Here are the codes from this page:

data forestg;
length SUBGP TEXT $14;
infile datalines delimiter=",";
input SUBGP $ IND N EPCT UPCT RR LCL UCL PVAL TVAR TEXT $ XP YP TID BAND;
datalines;
Age,0,.,.,.,.,.,.,0.004,1, ,.,.,1,.
< 13 Years,1,500,50.0,48.0,1.04,0.87,1.25,.,2, ,.,.,2,.
=> 13 Years,1,520,73.1,57.7,1.27,1.11,1.44,.,3, ,.,.,2,.
Sex,0,.,.,.,.,.,.,0.75,4, ,.,.,1,4
Female,1,500,54.4,49.2,1.11,0.93,1.31,.,5, ,.,.,2,5
Male,1,520,45.8,48.9,0.94,0.78,1.12,.,6, ,.,.,2,6
Ethnicity,0,.,.,.,.,.,.,0.35,7, ,.,.,1,.
Hispanic,1,80,60,46,1.30,0.86,1.98,.,8, ,.,.,2,.
Non-Hispanic,1,940,47.7,51.7,0.92,0.82,1.04,.,9, ,.,.,2,.
Length of Stay,0,.,.,.,.,.,.,0.0001,10, ,.,.,1,10
< 2 Days,1,484,50.4,57.9,0.87,0.74,1.03,.,11, ,.,.,2,11
=> 2 Days,1,536,41.5,57.4,0.72,0.61,0.86,.,12, ,.,.,2,12
Comorbidities,0,.,.,.,.,.,.,0.86,13, ,.,.,1,.
Yes,1,40,33.3,35.9,0.93,0.27,2.25,.,14, ,.,.,2,.
No,1,980,49.6,49.0,1.01,0.89,1.15,.,15, ,.,.,2,.
,.,.,.,.,.,.,.,.,16,Less Risk,0.4,16,2,.
,.,.,.,.,.,.,.,.,16,More Risk,1.8,16,2,.
;
run;

data forestmap2;
input ID $ value textcolor $ textsize textweight $;
datalines;
myid 1 black 7 bold
myid 2 black 5 normal
;
run;

proc sgplot data=forestg noautolegend dattrmap=forestmap2 ;
styleattrs axisextent=data;
refline BAND / lineattrs=(thickness=23 color=palegreen);
scatter x=RR y=TVAR / markerattrs=(symbol=squarefilled color=forestgreen)
xerrorlower=LCL xerrorupper=UCL errorbarattrs=(color=forestgreen);
scatter x=RR y=TVAR / markerattrs=(size=0) x2axis;
refline 1 / axis=x lineattrs=(color=forestgreen);
text x=XP y=TVAR text=TEXT / position=center;
yaxistable SUBGP / location=inside position=left textgroupid=myid textgroup=TID
labelattrs=(size=10 weight=bold) indentweight=IND nomissingchar label="Subgroup";
yaxistable N / location=inside position=left labelattrs=(size=10 weight=bold)
valueattrs=(size=7) nomissingchar label="Count";
yaxistable EPCT / location=inside position=right labelattrs=(size=10 weight=bold)
valueattrs=(size=7) nomissingchar label="Exposed";
yaxistable UPCT / location=inside position=right labelattrs=(size=10 weight=bold)
valueattrs=(size=7) nomissingchar label="Not Exposed";
yaxistable PVAL / location=inside position=right labelattrs=(size=10 weight=bold)
valueattrs=(size=7) nomissingchar label="P-Value";
yaxis reverse display=none colorbands=odd colorbandsattrs=(color=palegreen
transparency=0.5) offsetmin=0;
xaxis display=(nolabel) values=(0 to 2.5 by 0.5);
x2axis label="Risk Ratio" labelattrs=(size=10 weight=bold) display=(noline noticks
novalues);
run;

I would like to clip the comorbidity 'Yes' subgroup confidence interval [i.e. 0.27 - 2.25]

[ballardw · Accepted solution]

You don't say exactly which value you want to clip or what to. That would be a basic change of the value from the one you are currently plotting.

The possibilities with errorbars for changing only one value cap (which I think would be involved) would require an additional SCATTER statement with perhaps NOERRORBARCAPS and different X and Y variables so not confused with the other Scatter statements. Then either an additional Scatter with marker symbol like TriangleLeft or Triangleright, or the filled versions or even a GreaterThan or LessThan appearing at the position of the upper or lower error limit (i.e. where you want the arrow-like symbol to appear).

Curious as to why the concern though. The graph I get looks just fine. I'm not at all sure that the extra work is really worth much.

[Abimal_Zippi]

Thanks, @ballardw. And sorry, I wanted to clip the upperCL 2.25. You are right there is no problem with this graph. As I mentioned in my question, I am experimenting with this data to clip a value and add a marker symbol at the end (">".