Simple forest plot

[pink_poodle]

Hi,

Could someone please share SAS code to make a simple forest plot out of this (two studies go on the y axis):

data forest;                                                                                                                            
   input Study $1-16 OddsRatio LowerCL UpperCL;                                                                              
   format oddsratio lowercl uppercl 5.3;                                                                                                                                                                                       
datalines;                                                                                                                              
Modano  (1967)    0.590 0.096 3.634                                                                                              
Borodan (1981)    0.464 0.201 1.074                                                                                                                                                                                     
;   
run; 

Many thanks!

[Reeza · Accepted solution]

data forest;                                                                                                                            
   input Study $1-16 OddsRatio LowerCL UpperCL;                                                                              
   format oddsratio lowercl uppercl 5.3;                                                                                                                                                                                       
datalines;                                                                                                                              
Modano  (1967)    0.590 0.096 3.634                                                                                              
Borodan (1981)    0.464 0.201 1.074                                                                                                                                                                                     
;   
run; 


title "Odds Ratios with 95% Wald Confidence Limits";
proc sgplot data=forest noautolegend;
   scatter y=Study x=oddsratio / xerrorlower=LowerCL xerrorupper=UpperCL
           markerattrs=(symbol=diamondfilled);
   refline 1 / axis=x;
  * xaxis grid type=log label="Odds Ratio (log scale)";  /* specify log scale */
  * yaxis grid display=(nolabel) discreteorder=data reverse;
run;

Based on this post, but I commented out the log scale axis.

https://blogs.sas.com/content/iml/2015/07/29/or-plots-log-scale.html

Quite a few more complicated examples with data and code on the SAS blogs.

https://blogs.sas.com/content/graphicallyspeaking/2012/02/01/forest-plot-using-sgplot-procedure/

---

@pink_poodle wrote:

Hi,

Could someone please share SAS code to make a simple forest plot out of this (two studies go on the y axis):

data forest;                                                                                                                            
   input Study $1-16 OddsRatio LowerCL UpperCL;                                                                              
   format oddsratio lowercl uppercl 5.3;                                                                                                                                                                                       
datalines;                                                                                                                              
Modano  (1967)    0.590 0.096 3.634                                                                                              
Borodan (1981)    0.464 0.201 1.074                                                                                                                                                                                     
;   
run; 

Many thanks!

---

[pink_poodle]

@Reeza,

Thank you for a simple and elegant solution. I have just one follow-up question, I would like population numbers as a secondary y-axis on a graph like this (please also see image below):

Study                          N     

Modano (1967)          100

Borodan (1981)          200

How would you modify the code to accommodate the N variable?:

data forest;                                                                                                                            
   input Study $1-16 N OddsRatio LowerCL UpperCL;                                                                              
   format oddsratio lowercl uppercl 5.3;                                                                                                                                                                                       
datalines;                                                                                                                              
Modano  (1967)    100 0.590 0.096 3.634                                                                                              
Borodan (1981)    200 0.464 0.201 1.074                                                                                                                                                                                     
;   
run; 

title "Odds Ratios with 95% Wald Confidence Limits";
proc sgplot data=forest noautolegend;
   scatter y=Study x=oddsratio / xerrorlower=LowerCL xerrorupper=UpperCL
           markerattrs=(symbol=diamondfilled);
   refline 1 / axis=x;
  * xaxis grid type=log label="Odds Ratio (log scale)";  /* specify log scale */
  * yaxis grid display=(nolabel) discreteorder=data reverse;
run;

Many thanks!

[Reeza]

Several of the examples in the blog posts have this feature, have you tried using any of the code examples from there?

[pink_poodle]

I can put population numbers in columns to the right, but am not sure which part of code is responsible for those columns

(https://blogs.sas.com/content/graphicallyspeaking/files/2012/02/Full-SAS-93-Code.txt)

/*--Set up columns to create the stat tables--*/
  OR='OR'; LCL='LCL'; UCL='UCL'; WT='Weight';

  datalines;
Modano  (1967)    1  0.590 0.096 3.634  1
Borodan (1981)    1  0.464 0.201 1.074  3.5
Leighton (1972)   1  0.394 0.076 2.055  2
Novak   (1992)    1  0.490 0.088 2.737  2
Stawer  (1998)    1  1.250 0.479 3.261  3
Truark   (2002)   1  0.129 0.027 0.605  2.5
Fayney   (2005)   1  0.313 0.054 1.805  2
Modano  (1969)    1  0.429 0.070 2.620  2
Soloway (2000)    1  0.718 0.237 2.179  3
Adams   (1999)    1  0.143 0.082 0.250  4
Truark2  (2002)   1  0.129 0.027 0.605  2.5
Fayney2  (2005)   1  0.313 0.054 1.805  2
Modano2 (1969)    1  0.429 0.070 2.620  2
Soloway2(2000)    1  0.718 0.237 2.179  3
Adams2   (1999)   1  0.143 0.082 0.250  4
Overall           2  0.328 0.233 0.462  .
;
run;

[Reeza]

Then you try them one by one and see which does what.

[pink_poodle]

Sequential inactivation, yes :). I decided why not stretch a table next to a graph instead. It will look just as good.

[Reeza]

The code comments are pretty clear to me to be honest.

[pink_poodle]

@Reeza,
I have a quick follow-up question:
Is there a way to make xerrorlower and xerrorupper (i.e., odds ratio) values to appear above the tick marks?
Many thanks!

[pink_poodle]

*Here is one way of doing it;
title "Odds Ratios with 95% Wald Confidence Limits";
proc sgplot data=forest noautolegend;
   scatter y=Study x=oddsratio / xerrorlower=LowerCL xerrorupper=UpperCL
           markerattrs=(symbol=diamondfilled) datalabel=oddsratio
           datalabelattrs=(weight=bold);
   scatter y=study x=lowercl / datalabel = lowercl markerattrs=(size=0 color=crimson);
   scatter y=study x=uppercl / datalabel = uppercl markerattrs=(size=0 color=crimson);
   refline 1 / axis=x;
  * xaxis grid type=log label="Odds Ratio (log scale)";  
   /* specify log scale */
  * yaxis grid display=(nolabel) discreteorder=data reverse;
run;

[Reeza]

Change XERRORLOWER to YERRORLOWER?

I think that doesn't make sense though as scale is along the x axis so how would the Y axis also be the same scale?

[pink_poodle]

The output looks right:

[Reeza]

That shows error bars along the X axis only, no Y axis error bars.

[Reeza]

Update: Are you asking for datalabels for the error bars? If so, looks like you've got that solved.
If there isn't an option within Scatter you could use TEXT statements to plot those.