Caterpillar plots from random-effects proc glimmix output

[jeff_b]

Hi, I would appreciate a help with creating a caterpillar plot from the generalized linear mixed (binomial distribution) models with random intercepts output. I came across with an article on creating a caterpillar plot from proc mcmc output using the cater macro:

ods graphics on;
%CATER(data=postout, var=delta: );
ods graphics off;

I want to do a similar plot but with the proc glimmix output (SolutionR=random effect estimates

and plotting the ‘wiskers’ as +/- 1.96 x the standard errors).

[Rick_SAS]

If you post your code, we can be more helpful. What statistic are you trying the plot? The parameter estimates and their standard errors?

If so, see this article about regression coefficient plots.

As shown in the article, the basic idea is to output the information you want to plot (estimates and stderrs) to a SAS data set. then use a DATA step to create two variables: Lower=Estimate-1.96*StdErr and Upper=Estimate-+.96*StdErr.

Use the SCATTER statement in PROC SGPLOT and the XERRRORLOW= and XERRORUPPER= options to display the 1.96*StdErr intervals.

Other articles that cerate similar plots are 

"Coverage probability of confidence intervals"

and

"Odds ratio plots"

[jeff_b]

Thank you Rick! Your response is extremely helpful.

Posting my code below:

data temp;
set SolutionR;

 Lower=Estimate-1.96*StdErrPred;
 Upper=Estimate+1.96*StdErrPred;
run;

proc sort data=temp;  by descending Estimate;  run;

proc sgplot data=temp noautolegend;
   scatter y=Estimate x=Subject  / yerrorlower=Lower yerrorupper=Upper  markerattrs=(symbol=diamondfilled);
   refline 0 / axis=y;
   xaxis grid;
   yaxis grid display=(nolabel) discreteorder=data reverse;
run;

[Rick_SAS · Accepted solution]

Great. Glad you solved your problem. If you have other questions, please post them. Otherwise, please mark the thread as solved.