topic Re: Predicted probability graphic fitted curve/confidence interval. in Graphics Programming

Predicted probability graphic fitted curve/confidence interval.

jesspurse — Fri, 03 Apr 2020 00:08:38 GMT

I've been told I need to have just the fitted curve and confidence interval in my graphic, not all the data points. I have literally no idea how to do it.

This is my code to make the predicted probablity graphic, and it's attached as a pdf.

proc glm data=teen.obdata;
weight p_teenpreg;
model p_adverse=minroadmiles miles_sq rural rural*miles_sq;
store intmodel;
run;
quit;

proc plm source=intmodel;
score data=teen.obdata out=_temp1_ predicted / ilink;
run;
ods listing gpath="\\Mac\Home\Desktop";
ods graphics / imagename="small" imagefmt=png;
proc sgplot data = _temp1_;
series x = minroadmiles y = predicted / group = rural;
yaxis label = "Predicted Probability of Adverse Birth Outcomes";
xaxis values= (0 5 10 15 20 25 30 35 40) label = "Miles";
title “Predicted Probability by Road Miles”;
run;
title;
ods graphics off;

Re: Predicted probability graphic fitted curve/confidence interval.

DanH_sas — Fri, 03 Apr 2020 02:32:19 GMT

Can you post the output of "proc contents data=_temp1_; run;"?

Re: Predicted probability graphic fitted curve/confidence interval.

jesspurse — Fri, 03 Apr 2020 03:01:39 GMT

attached!

Re: Predicted probability graphic fitted curve/confidence interval.

Rick_SAS — Tue, 07 Apr 2020 20:51:21 GMT

The problem is that your model has covariates, so plotting the predicted values against one of the covariates (minroadmiles) does not result in a smooth curve. You can use the EFFECTPLOT statement in PROC PLM to produce a sliced fit plot, which enables you to visualize a slice through the regression surface. For more discussion and examples, see Visualize multivariate regression models by slicing continuous variables.

Also see, "3 ways to add confidence limits to regression curves in SAS."