The scatterplot in section 7.4 is a plot we could use repeatedly. We demonstrate how to create a macro (SAS, section A.8) and a function (R, section B.5) to do it more easily.

SAS

%macro logiplot(x=x, y=y, data=, jitterwidth=.05, smooth=50);
data lp1;
set &data;
if &y eq 0 or &y eq 1;
jitter = uniform(0) * &jitterwidth;
if &y eq 1 then yplot = &y - jitter;
else if &y eq 0 then yplot = &y + jitter;
run;

axis1 minor=none label = ("&x");
axis2 minor=none label = (angle = 270 rotate = 90 "&y");
symbol1 i=sm&smooth.s v=none c=blue;
symbol2 i=none v=dot h=.2 c=blue;
proc gplot data=lp1;
plot (&y yplot) * &x / overlay haxis=axis1 vaxis=axis2;
run;
quit;

R

logiplot <- function(x, y, jitamount=.01, smoothspan=2/3, 
   xlabel="X label", ylabel="Y label") {
   jittery <- jitter(y, amount=jitamount/2)
   correction <- ifelse(y==0, jitamount/2, -jitamount/2)
   jittery <- jittery + correction
   plot(x, y, type="n", xlab=xlabel, ylab=ylabel)
   points(x, jittery, pch=20, col="blue")
   lines(lowess(x, y, f=smoothspan), col="blue")
}

We’ll load the example data set from the book via the web (section 1.1.6), then make a plot of the real data.

SAS

filename myurl 
   url 'http://www.math.smith.edu/sasr/datasets/help.csv'  lrecl=704;

proc import
   datafile=myurl
   out=ds dbms=dlm;
   delimiter=',';
   getnames=yes;
run;

%logiplot(x = mcs, y = homeless, data=ds, smooth=40);

R

ds <- read.csv("http://www.math.smith.edu/sasr/datasets/help.csv")
logiplot(ds$mcs, ds$homeless, jitamount=.05, smoothspan=.3, 
   xlabel="mcs", ylabel="homeless")

The resulting plots are quite similar, but still differ with respect to the smoother and the exact jitter applied to each point.



(Click for a bigger picture.)