LD Statistic (Virtual Base Class)

Abstract container for pre-computed LD statistics. Subclasses provide method-specific representations: eigendecompositions (for LDER/HDL/sHDL) and LD score matrices (for S-LDSC/g-LDSC).

An LdStatistic is a GRanges of the reference variants: one range per SNP, with SNP / A1 / A2 (and optionally MAF) in mcols. Per-variant statistics live in mcols too, so subsetting narrows the ranges and the statistics together. The genome build comes from seqinfo() rather than a slot of its own.

A statistic built from a genotype panel (or a GenotypeMixture) keeps that reference rather than its blocks' LD: each block's LD is read from the panel when an estimator reaches the block, so only one block's is ever in memory, and mcols() records each variant's row in the panel as panelRow. One built from LdMatrix blocks stores what it needs of their LD instead.

Slots

ldBlocks

A GRanges of LD block intervals.

nSamples

Integer, sample size of the LD reference panel.

inSample

Logical, whether the LD reference is from the same cohort as the GWAS (affects bias correction).

ldReference

The genotype panel or GenotypeMixture the statistic reads its blocks' LD from, or NULL when it was built from LdMatrix blocks.