DM Stat-1

Rare Event Sampling

Bruce Ratner, Ph.D.

DATA sample_data;
do j = 1 to 1500000;
if j le 10000 then RESPONSE=1;
else RESPONSE=0;
output;
end;
RUN;

title1' sample_data RESPONSE rate ';
PROC FREQ; table RESPONSE;
RUN;
PROC MEANS data=sample_data;
var RESPONSE;
output out=sam_CNTS (drop=_TYPE_ _FREQ_)
mean=pop_RespRate sum=sam_RR;
RUN;

DATA sam_CNTS;
set sam_CNTS;
m=1;
RUN;

DATA nr rr;
set sample_data;
uni=uniform(12345);
m=1;
if RESPONSE=0 then output nr;
if RESPONSE=1 then output rr;
RUN;

PROC SORT data=nr ;
by uni;
RUN;

DATA nr;
merge nr sam_CNTS;by m;
sam_NR=sam_RR*5;
if _n_ le (sam_NR);
RUN;

DATA nr_rr;
set nr rr;
pop_NRespRate=1-pop_RespRate;
_r=compress(put(pop_RespRate,8.7),'.');
_n=compress(put(pop_NRespRate,8.7),'.');
pop_RR=input(_r,10.);
pop_NR=input(_n,10.);
m=1;
RUN;

DATA sample_down;
merge nr_rr nr;
by m;
WT=1;
if RESPONSE=0
then WT=int( (sam_RR/sam_NR)/(pop_RR/pop_NR) );
RUN;

title1' sample_down RESPONSE rate';
PROC MEANS data=sample_down;var RESPONSE;
freq WT;
RUN;

For more information about this article, call Bruce Ratner at 516.791.3544 or
1 800 DM STAT-1; or e-mail at br@dmstat1.com.