## 13 March 2015
                                                                                                                                                                                                                                                                                                                                                                   
## precision for estimating the proportion
z <- qnorm(.975)
p <- .5
N <- 1:200
D1 <- sqrt(z^2*p*(1-p)/N)
D1


plot(N,D1,type='l')


## roughly quadruple sample size to decrease half-length of confidence interval by half
plot(N,D1,type='l',xlim=c(0,50))
N[round(D1,3)==.200]
N[round(D1,3)==.100]