Module voom 

Mean-variance modelling of count data at the observation level. Port of limma’s voom (voom.R) and the span heuristic chooseLowessSpan.

voom transforms counts to log2-counts-per-million, fits a gene-wise linear model, fits a LOWESS trend of the residual standard deviation against the average log-count, and converts that trend into per-observation precision weights for entry to the usual lmFit -> eBayes pipeline.

The current port covers the default path: an optional design and library sizes, normalize.method = "none", no offsets and no blocking/correlation. Per-array prior weights are supported via [voom_weighted], which backs voom_with_quality_weights (voom + gene-by-gene array quality weights).

Structs

VoomOutput

Result of voom: the log-CPM matrix E, the per-observation precision weights, the design used, the per-sample library sizes, and the span.

VoomQualityWeights

Result of voom_with_quality_weights: a VoomOutput-style payload whose weights already fold in the per-sample sample_weights.

VoomaByGroupOutput

Result of vooma_by_group.

VoomaLmFit

Result of vooma_lm_fit.

VoomaOutput

Result of vooma: the per-observation precision weights, the design used, and the span.

Functions

choose_lowess_span

Choose a LOWESS span as a function of the number of points. Port of chooseLowessSpan: larger spans for small datasets, smaller for large ones.

voom

voom(counts, design, lib.size, span, adaptive.span) with normalize.method = "none".

voom_with_quality_weights

voomWithQualityWeights(counts, design, span, adaptive.span, var.design, method="genebygene") — combine voom observation weights with array quality weights estimated gene-by-gene.

vooma

vooma(y, design, span, legacy.span) — the continuous-data analog of voom, with no predictor and no blocking.

vooma_by_group

voomaByGroup(y, group, design, span, legacy.span) — fit a separate vooma mean-variance trend within each level of group and assemble the per-group observation weights into one n_genes x n_samples matrix for entry to lmFit.

vooma_lm_fit

voomaLmFit(y, design, prior.weights, span, legacy.span) — vooma followed by a weighted lmFit, returning a fit ready for eBayes.