PROC LOGISTIC to handle Randomized Complete Block Design (RCBD)

The main idea of blocking is to ensure that the random sample you draw from the population of interest does not, by random chance, have some undesirable properties. 

*Use risk scores (riskScr) and response groups (respGrp) to create blocks;
data blocks;
set s.dem0303;
if riskScr < 650 then riskGrp=1;
else if riskScr < 750 then riskGrp=2;
else riskGrp =3;
if respScr < 260 then respGrp=1;
else if respScr < 280 then respGrp=2;
else respGrp =3;
block = compress(riskGrp||respGrp);
random=ranuni(8675309);
run;

proc sort data=blocks;
by block;
run;

*randomly assigned units to treatments within each block;
proc rank data=blocks out=blocks groups=8;
by block;
var random;
ranks treatment;
run;

*Generate design matrix (factors: interestRate sticker priceGraphic);
data blocks;
set blocks;
array factors(3) interestRate sticker priceGraphic;
do i = 1 to dim(factors);
factors(i)=substr(put(treatment,binary3.),i,1);
end;
run;

*Analysis after collecting the data;
*Block | InterestRate | Sticker | PriceGraphic;
ods select globalTests type3;
title "Block | InterestRate | Sticker | PriceGraphic";
proc logistic data=s.dem0304 namelen=40;
class block interestRate sticker priceGraphic;
model response(event="1")=
block
interestRate|sticker|priceGraphic @1
interestRate|sticker|priceGraphic @2
interestRate|sticker|priceGraphic;
run;

Note:
It is possible to get the same parameter estimates in a slightly easier to read format. The above code uses the '@' notation in the MODEL statement to arrange the terms in the model by their polynomial order. A|B|C|D in a MODEL statement yields every term from the main effects up to the three-factor interaction.