Skip to contents

Build the selection portion of a spec(). This helper configures the current correlation-based edge-selection path by choosing the correlation method used to relate each edge to the outcome and how the associated level is interpreted.

Usage

cpm_selection_cor(
  method = c("pearson", "spearman"),
  criterion = c("p_value", "absolute", "proportion"),
  level = 0.01
)

Arguments

method

Correlation method used during edge selection. "pearson" is the default CPM path; "spearman" can be useful when a monotonic but non-linear relationship is expected.

criterion

How level is interpreted. "p_value" thresholds the absolute correlation by the critical value implied by level. When covariates are used, the test accounts for the effective training-set covariate rank. Spearman screening uses the same t approximation after residualization. "absolute" treats level as a direct absolute-correlation cutoff. "proportion" retains a per-sign proportion of positive and negative edges separately. Because edge counts are discrete and ties at the cutoff are retained, the realized proportion may be slightly larger than the requested level.

level

Numeric level associated with criterion.

Examples

cpm_selection_cor()
#> $type
#> [1] "cor"
#> 
#> $method
#> [1] "pearson"
#> 
#> $criterion
#> [1] "p_value"
#> 
#> $level
#> [1] 0.01
#> 
#> attr(,"class")
#> [1] "cpm_selection_spec"
cpm_selection_cor(method = "spearman", criterion = "absolute", level = 0.1)
#> $type
#> [1] "cor"
#> 
#> $method
#> [1] "spearman"
#> 
#> $criterion
#> [1] "absolute"
#> 
#> $level
#> [1] 0.1
#> 
#> attr(,"class")
#> [1] "cpm_selection_spec"