Skip to contents

Non-parametric (permutation) SPM

Source: R/stats-spm-nonparam.R
spmSnPM.Rd

Builds the permutation distribution of the field-maximum statistic to obtain a distribution-free critical threshold with strong FWER control (Nichols & Holmes 2002), plus permutation cluster- and set-level p-values. This is the validity cross-check for the parametric RFT thresholds: as the field smoothness grows the permutation threshold converges to the RFT threshold.

Usage

spmSnPM(
  x,
  groups = NULL,
  group1 = NULL,
  group2 = NULL,
  statistic = c("t", "F", "T2"),
  n_permutations = 1000,
  alpha = 0.05,
  two_tailed = TRUE,
  vector_components = NULL,
  seed = NULL
)

Arguments

x

A PhysioExperiment/matrix (time x obs) for t/F, or a 3D array / list of component matrices for T2.

groups

Grouping factor (one per observation) for F or T2, or a two-sample t.

group1, group2

Observation indices for a two-sample t (group2 = NULL gives a one-sample sign-flip test).

statistic

"t", "F", or "T2".

n_permutations

Number of random permutations (default 1000).

alpha

Significance level (default 0.05).

two_tailed

For t: use the two-tailed field maximum max|t|.

vector_components

Passed to the T2 array coercion.

seed

Optional RNG seed for reproducibility.

Value

A list of class "spm_result" (test_type = "snpm") with the observed statistic field, the permutation threshold, clusters with permutation p-values, pointwise permutation p_values, and the perm_max distribution.

References

Nichols & Holmes 2002; Pataky 2016. spm1d.stats.nonparam.

See also

spmTTest(), spmAnova(), spmMANOVA()

Examples

set.seed(1)
g1 <- matrix(rnorm(50 * 10), 50); g2 <- matrix(rnorm(50 * 10), 50)
g2[20:30, ] <- g2[20:30, ] + 1.5
spmSnPM(cbind(g1, g2), group1 = 1:10, group2 = 11:20,
        n_permutations = 200, seed = 1)
#> SPM Analysis Result
#> ==================
#> Test type: snpm
#> Time points: 50
#> Alpha: 0.050
#> Threshold: 4.125
#> Permutations: 200
#> 
#> Significant clusters: 2
#>   Cluster 1: [23-23] extent=1, p=0.0050
#>   Cluster 2: [28-28] extent=1, p=0.0050