data original;
    do j = 1 to 50;
        rnorm = 22 + 5*rannor(345768);
        output;
    end;
run;

data _NULL_;
    if 0 then set original nobs=n;
    call symputx('nrows',n);
    stop;
run;

%put nobs=&nrowsd;
data bootsamp;
do sampnum = 1 to 1000; /* Can use different number of reps */
    do i = 1 to numrecs; /* Want same no. obs as in ORIGINAL */
        x = round(ranuni(0)* &nrows); /*x randomly selected from values */
        set original nobs = numrecs point = x; /* This selected the xth output; /* Send the selected observation to the new data set */
    end;
end;
stop; /* Required when using point */
run;
/*proc print data=bootsamp; run; */

data bootorig;
set original (in=a) bootsamp (drop=i);
if a then sampnum = 0;
run;
/* proc print data=bootorig; run; */

proc means noprint data=bootorig;
class sampnum;
var rnorm;
output out=bootout
    mean=mean
    var=var
    n=n;
run;

data calcbias;
    merge bootout;
    by sampnum;
        sampmean = mean;
        sampvar = var;
    retain origmean;
    if sampnum=0 then origmean=mean; /* Calculate bias and MSE */
bias = sampmean-origmean;
bias2 = bias*bias;
mse = sampvar - bias2;
run;

proc print data=calcbias;
var bias mse;
run;