est_sv: Estimate the surrogate variables.
In seqgendiff: RNA-Seq Generation/Modification for Simulation

est_sv

R Documentation

Estimate the surrogate variables.

Description

This will use either sva or an SVD on the residuals of a regression of mat on design_obs to estimate the surrogate variables.

Usage

est_sv(mat, n_sv, design_obs, use_sva = FALSE)

Arguments

`mat`	A numeric matrix of RNA-seq counts. The rows index the genes and the columns index the samples.
`n_sv`	The number of surrogate variables.
`design_obs`	A numeric matrix of observed covariates that are NOT to be a part of the signal generating process. Only used in estimating the surrogate variables (if `target_cor` is not `NULL`). The intercept should not be included (it will sometimes produce an error if it is included).
`use_sva`	A logical. Should we use surrogate variable analysis (Leek and Storey, 2008) using `design_obs` to estimate the hidden covariates (`TRUE`) or should we just do an SVD on `log2(mat + 0.5)` after regressing out `design_obs` (`FALSE`)? Setting this to `TRUE` allows the surrogate variables to be correlated with the observed covariates, while setting this to `FALSE` assumes that the surrogate variables are orthogonal to the observed covariates. This option only matters if `design_obs` is not `NULL`. Defaults to `FALSE`.

Value

A matrix of estimated surrogate variables. The columns index the surrogate variables and the rows index the individuals. The surrogate variables are centered and scaled to have mean 0 and variance 1.