projecting regression line beyond data

[linlin87]

Hi SAS Forus,

I have a regression line and it only creates this for the x data that I have. How do I get to project beyond the x data?

proc sgplot data=input;
reg x=months y=change_height / clm cli;
series x=months y=change_height /group=ani markerattrs=(symbol=CircleFilled);
scatter x=months y=change_height /group=ani markerattrs=(symbol=CircleFilled);
xaxis max=20;
yaxis min=-20 max=20;
refline -10 10 / axis=y;
run;

Would this be easier to implement with proc reg?

Thanks!

[ballardw]

My approach would be to add the additional values into the base data for the independent variable(s), x axis,and missing values for the dependent variable(s), y-axis, and use an appropriate regression procedure to create predicted values and use the resulting data set to plot such extrapolated values. Recommend including the estimated confidence limits as well.

[PaigeMiller · Accepted solution]

What @ballardw is referring to (I think) is the "Missing Response Trick", at least that's what it is called in this document (https://www.sas.com/content/dam/SAS/support/en/sas-global-forum-proceedings/2019/3337-2019.pdf) and if you scroll down to the section entitled "Scoring Data the Old-Fashioned Way", you will find code to do this.

Since @Rick_SAS has written blog posts on everything you might ever want to do in SAS, here's his blog post on the subject.

https://blogs.sas.com/content/iml/2014/02/17/the-missing-value-trick-for-scoring-a-regression-model....

[ballardw]

---

@PaigeMiller wrote:

What @ballardw is referring to (I think) is the "Missing Response Trick", at least that's what it is called in this document (https://www.sas.com/content/dam/SAS/support/en/sas-global-forum-proceedings/2019/3337-2019.pdf) and if you scroll down to the section entitled "Scoring Data the Old-Fashioned Way", you will find code to do this.

Since @Rick_SAS has written blog posts on everything you might ever want to do in SAS, here's his blog post on the subject.

https://blogs.sas.com/content/iml/2014/02/17/the-missing-value-trick-for-scoring-a-regression-model....

---

Learned that back about 1986, so I guess it qualifies as "old fashioned". At least we don't really have to use Proc Plot to show results... (Really, not GPlot, not Sgplot but line-printer text based Proc Plot).