samplesize.bin {Hmisc} | R Documentation |
Computes sample size(s) for 2-sample binomial problem given vector or scalar probabilities in the two groups.
samplesize.bin(alpha, beta, pit, pic, rho=0.5)
alpha |
scalar ONE-SIDED test size, or two-sided size/2 |
beta |
scalar or vector of powers |
pit |
hypothesized treatment probability of success |
pic |
hypothesized control probability of success |
rho |
proportion of the sample devoted to treated group (0 <rho < 1) |
TOTAL sample size(s)
Rick Chappell
Dept. of Statistics and Human Oncology
University of Wisconsin at Madison
chappell@stat.wisc.edu
alpha <- .05 beta <- c(.70,.80,.90,.95) # N1 is a matrix of total sample sizes whose # rows vary by hypothesized treatment success probability and # columns vary by power # See Meinert's book for formulae. N1 <- samplesize.bin(alpha, beta, pit=.55, pic=.5) N1 <- rbind(N1, samplesize.bin(alpha, beta, pit=.60, pic=.5)) N1 <- rbind(N1, samplesize.bin(alpha, beta, pit=.65, pic=.5)) N1 <- rbind(N1, samplesize.bin(alpha, beta, pit=.70, pic=.5)) attr(N1,"dimnames") <- NULL #Accounting for 5% noncompliance in the treated group inflation <- (1/.95)**2 print(round(N1*inflation+.5,0))