genodds (version 1.1.2)

genodds.power: Power Calculations for Generalized Odds Ratios

Description

Provides power analysis for Agresti's Generalized Odds Ratios.

Usage

genodds.power(
  p0,
  p1,
  N = NULL,
  power = NULL,
  alpha = 0.05,
  ties = "split",
  w = c(0.5, 0.5),
  direction = "two.sided"
)

Arguments

p0

A numeric vector contianing the probabilities in control group.

p1

A numeric vector contianing the probabilities in treatment group.

N

A numeirc vector containing total sample sizes.

power

A numeric vector containing required total sample size.

alpha

Type 1 error.

ties

A string specifying how ties should be treated. Should be equal to "split" 0.5 for WMW Odds, or "drop" for Agresti's GenOR.

w

A numeric vector of length 2 specifying the relative weighting of sample size between treatment groups.

direction

Direction for hypothesis test. Must be one of "two.sided","upper.tail" or "lower.tail".

Value

If power is supplied

A numeric vector containing required sample sizes to achieve specified powers.

If N is supplied

A numeric vector containing power at specified sample sizes.

Details

See genodds for explanation of generalized odds ratios.

N provides the total sample size. Sample size per group can be calculated by N*w/sum(w).

When power is supplied, if no sufficient sample size is found then this function will return Inf.

References

O'Brien, R. G., & Castelloe, J. (2006, March). Exploiting the link between the Wilcoxon-Mann-Whitney test and a simple odds statistic. In Thirty-first Annual SAS Users Group International Conference.

Examples

# NOT RUN {
# Provide theoretical distributions of outcomes for each group
# Distributions taken from Lees et. al. (2010). See ?alteplase for a citation.
p0 <- c(0.224,0.191,0.082,0.133,0.136,0.043,0.191)
p1 <- c(0.109,0.199,0.109,0.120,0.194,0.070,0.200)

# Calculate sample size required to achieve 80% and 90%
# power for these distributions
genodds.power(p0,p1,power=c(0.8,0.9))

# genodds.power suggests a total sample size of 619 for 80% power.
# Round up to 620 for even sample size per group

# Confirm these sample sizes lead to 80% and 90% power
genodds.power(p0,p1,N=c(620,830))

# }