Power for t tests, including arbitrary linear hypotheses

Usage

tpower(
  n,
  delta,
  lam = c(1, -1),
  b,
  sd = 1,
  alpha = 0.05,
  w = rep(1, g),
  n1,
  n2,
  verbose = (N == 1)
)

tsamsize(delta, b = c(delta, 0), w = rep(1, g), power = 0.8, upper = 5000, ...)

Arguments

n: Sample size (total).
delta: Effect size; if specified, a two-group design is assumed. If more than two groups, use b instead.
lam: Contrast.
b: If a vector, the coefficient for each group. If a matrix, must have length(n) rows.
sd: Standard deviation of outcome.
alpha: Type I error rate.
w: Weights for unequal allocation (normalized to 1).
n1, n2: Sample size for group 1, 2.
verbose: Print details for power calculations. Default: TRUE unless vectorized.
power: Desired power.
upper: Upper bound for tsamsize(); increase if tsamsize hits this bound. Default: 5000.
...: For tsamsize(), additional arguments to be passed to tpower().

Examples

tpower(100, 0.5)
#> Group 1: 50
#> Group 2: 50
#> Power:   0.697
tpower(10*(6:9), 0.5)                # Vectorize sample size
#> [1] 0.4778410 0.5406591 0.5981316 0.6501772
tpower(100, seq(0.25, 1, by=0.25))   # Vectorize effect size
#> [1] 0.2350874 0.6968888 0.9602024 0.9986074
tpower(100, 0.5, alpha=seq(0.01, 0.05, 0.01))   # Vectorize alpha
#> [1] 0.4529913 0.5553191 0.6178457 0.6624989 0.6968888
tpower(99, b=c(0.5, 0.5, 0), lam=c(1,1,-1))    # A multi-group example
#> Group 1: 33
#> Group 2: 33
#> Group 3: 33
#> Power:   0.907
tsamsize(0.5)
#> n1 n2 
#> 64 64 
tsamsize(lam=c(-1,1,0,0), b=c(0,1,1,1))
#> n1 n2 n3 n4 
#> 17 17 17 17