<PRE><B>Question 1110326</B>
Within 4% and confidence 99%
the error 0.04=z* sqrt (SE), where z=2.576 and SE is sqrt (p*(1-p))/n.  Use 0.5 when no estimate is available, since that is the most conservative (largest) product of two decimals whose sum is 1.
2.576*sqrt(0.25)/sqrt (n) equals error
2.576*0.5/sqrt (n)=0.04
1.288/0.04=sqrt(n); square both sides
n=1036.84 or 1037.

With a known fraction of 0.18, p*(1-p)=0.1476
(1.288/0.1476)^2=76.15 or 77
When a prior estimate is available and is significantly different from 0.5, the sample size falls dramatically.


</PRE>