poolcushion_t function

Visualize T-test Power for Pooling Design as Function of Processing Error Variance

Useful for choosing a sample size such that power will be adequate even if the processing errors are larger than anticipated.

poolcushion_t(g = NULL, n = NULL, d = NULL, mu1 = NULL,
  mu2 = NULL, sigsq = NULL, sigsq1 = sigsq, sigsq2 = sigsq,
  sigsq_p_predicted = 0, sigsq_p_range = NULL, sigsq_m = 0,
  multiplicative = FALSE, alpha = 0.05, beta = 0.2, labels = TRUE)

Arguments

• g: Numeric value specifying the pool size.
• n: Numeric value specifying the number of assays per group. If unspecified, function figures out n required for 100 (1 - beta)% power when sigsq_p = 0.
• d: Numeric value specifying true difference in group means.
• mu1, mu2: Numeric value specifying group means. Required if multiplicative = TRUE.
• sigsq: Numeric value specifying the variance of observations.
• sigsq1, sigsq2: Numeric value specifying the variance of observations for each group.
• sigsq_p_predicted: Numeric value specifying predicted processing error variance. Used to calculate n if n is unspecified.
• sigsq_p_range: Numeric vector specifying range of processing error variances to consider.
• sigsq_m: Numeric value specifying the variance of measurement errors.
• multiplicative: Logical value for whether to assume multiplicative rather than additive errors.
• alpha: Numeric value specifying type-1 error rate.
• beta: Numeric value specifying type-2 error rate. Only used if n = NULL.
• labels: Logical value.

Returns

Plot generated by ggplot.

Examples

# Determine optimal pool size and number of assays to detect a difference in
# group means of 0.5, with a common variance of 1, processing errors with
# variance of 0.1, and measurement errors with variance of 0.2. Assume costs
# of $100 per assay and $10 per subject.
poolcost_t(
  g = 1: 10,
  d = 0.5,
  sigsq = 1,
  sigsq_p = 0.1,
  sigsq_m = 0.2,
  assay_cost = 100,
  other_costs = 10
)

# Visualize how power of the study will be affected if the true processing
# error variance is not exactly 0.1.
poolcushion_t(
  g = 7,
  n = 29,
  d = 0.5,
  sigsq = 1,
  sigsq_p_predicted = 0.1,
  sigsq_m = 0.2
)