Per-Effect Retained Posterior Mass

The share of each single effect's posterior that survived a variant subset – the diagnostic that says how much a reconciliation cost.

Usage

getRetainedMass(x, ...)

# S4 method for class 'FineMappingResultBase'
getRetainedMass(x, ...)

Arguments

x

A fine-mapping collection.

...

Additional arguments passed to methods.

Value

A data.frame, one row per (element, effect), with the element's identity columns, the effect index, and its retained mass.

Details

Reported always, never used as a filter. The mass is continuous, not bimodal, once the causal variant is untyped (median 0.78 at 30 and 0.9), so there is no natural cut point and no default drop threshold is imposed.

susieR's own V -> 0 null rule is not a substitute: across 135 effects it fired zero times, including all 52 whose retained mass was ~0 – retained variants in LD with a dropped causal variant still beat the null. An explicit diagnostic is the only thing that surfaces this.

Examples

data(qtlFineMappingExample)
data(gwasFineMappingExample)
# Nothing has been reconciled yet, so there is no mass to report.
nrow(getRetainedMass(qtlFineMappingExample))
#> [1] 0
both <- intersectVariants(qtlFineMappingExample, gwasFineMappingExample)
getRetainedMass(both$x)
#> # A tibble: 2 × 7
#>   study   context   trait  method effect retainedMass nVariants
#>   <chr>   <chr>     <chr>  <chr>   <int>        <dbl>     <int>
#> 1 study_1 context_1 gene_1 susie       1        1.000      2828
#> 2 study_1 context_1 gene_1 susie       2        1.000      2828