02-27-2016 10:19 AM
Hi, I need a diagram to show covariance/correlation attributes of my bivariate data. My idea was a scatter plot which is proportional for both variables (1 unit on the x-axis has the same lengh as 1 unit on the y-axis) and contains a line for each mean and one for the regression line. This way the degree in which extreme values of x are associated with extreme values of y could be graphically assessed.
A scatter plot with a regression line is no problem in SAS but I couldn't find options to make the diagram proportional and to plot lines for the means.
02-27-2016 12:13 PM - edited 02-27-2016 12:15 PM
What you need is what is known as an EQUATED container in GTL. SGPLOT scripts out the LAYOUT OVERLAY to create the graph. What you can do is use the TMPLOUT="filename" option on the SGPLOT statement to get the GTL Template scripted by SGPLOT. Edit this program, replace LAYOUT OVERLAY with LAYOUT OVERLAYEQUATED. Then, compile the template, and run proc SGRENDER with your data and this new template.
define statgraph sgplot;
layout OVERLAYEQUATED / <options>;
<keep these statements>
proc sgrender data=<your data> template=sgplot;
Depending on the program, you may need to remove / replace some of the options on the LAYOUT statement.
02-27-2016 12:45 PM
You can also use the ASPECT= option on the PROC SGPLOT statement. For a comparison of the various options, see the article "Preserving the aspect ratio of th data in ODS graphics."