Summarize a cpm_resamples object — summary.cpm

Summarize a cpm_resamples object

Usage

# S3 method for class 'cpm_resamples'
summary(object, ...)

# S3 method for class 'cpm_resamples_summary'
print(x, ...)

Arguments

object: An object of class cpm_resamples.
...: For future extension. Currently ignored.
x: An object of class cpm_resamples_summary.

Value

A cpm_resamples_summary object with the following elements:

metrics: A data frame with columns level, metric, prediction, estimate, std_error, and method. Resample summaries store pooled errors and pooled correlations at level = "pooled", and fold-wise correlation summaries at level = "foldwise".
edges: Aggregated edge-selection rates, or NULL when edges were not stored.
params: A list containing summary-relevant resampling settings.

Details

summary.cpm_resamples() is designed to give a compact default report. It leads with pooled out-of-fold error metrics (RMSE and MAE), then reports pooled and fold-wise correlations as supplementary statistics. This keeps the default summary usable even when fold-wise correlations are undefined for some resampling schemes, such as leave-one-out resampling.

Examples

withr::local_seed(123)
conmat <- matrix(rnorm(200), nrow = 20)
behav <- rowMeans(conmat[, 1:5, drop = FALSE]) + rnorm(20, sd = 0.2)
res <- fit_resamples(cpm_spec(), conmat = conmat, behav = behav, kfolds = 4)

summary(res)
#> CPM resample summary:
#>   Number of folds: 4
#>   Prediction error:
#>     RMSE:
#>       Combined: 0.432
#>       Positive: 0.432
#>       Negative: 0.452
#>     MAE:
#>       Combined: 0.365
#>       Positive: 0.365
#>       Negative: 0.358
#>   Pooled correlations (Pearson):
#>     Combined: 0.243
#>     Positive: 0.243
#>     Negative: -0.195
#>   Fold-wise correlations (Pearson):
#>     Combined: 0.454
#>     Positive: 0.454
#>     Negative: NA