Alpha adjustment function: SAFFRON — alpha

These function accept a vector of p-values and vector indicating batch of p-values It produces a vector of alpha values.

Usage

alpha_saffron(
  pval,
  batch,
  nodesize,
  thealpha = 0.05,
  thew0 = 0.05 - 0.001,
  depth = NULL
)

Arguments

pval: Is a numeric vector of p-values
batch: A character or factor or even numeric vector indicating grouping among the p-values and the order of the p-values. The first element pval and first element of batch should be the first p-value produced in the tree and the subsequent p-values should be produced on the children of that root (or children of the children of the root, etc.)
nodesize: Is vector indicating the information used to create each p-value. For example, it could be the number of observations, or a weighted version.
thealpha: Is the overall error rate for a given test
thew0: Is the starting "wealth" of the alpha investing procedure
depth: Integer vector of tree depths for each p-value (1 = root). Accepted for interface compatibility with tree-structured alpha adjusters but not used by this function.

Value

A vector of alpha values

Examples

# Example with SAFFRON procedure for hierarchical testing
pvals <- c(0.01, 0.04, 0.12, 0.08, 0.15, 0.02)
batches <- c(1, 2, 2, 3, 3, 3) # Tree depth levels
node_sizes <- c(100, 50, 50, 25, 25, 25) # Sample sizes at each node

# Apply SAFFRON procedure
alpha_vals <- alpha_saffron(pvals, batches, node_sizes, thealpha = 0.05)

# Compare p-values to adjusted alpha levels
data.frame(
  pval = pvals,
  batch = batches,
  alpha = alpha_vals,
  significant = pvals <= alpha_vals
)
#>   pval batch      alpha significant
#> 1 0.01     1 0.01714961        TRUE
#> 2 0.04     2 0.01749961       FALSE
#> 3 0.12     2 0.01749961       FALSE
#> 4 0.08     3 0.01749961       FALSE
#> 5 0.15     3 0.01749961       FALSE
#> 6 0.02     3 0.01749961       FALSE