glsSymm: Fit General Linear Model with 'corSymm' Correlation Structure...

In ntyet/rmRNAseq: RNA-Seq Analysis for Repeated-Measures Data

Fit General Linear Model with corSymm Correlation Structure for One Gene

Description

This function fits gls model with REML estimation method, corSymm unstructured correlation for one gene in a RNA-seq repeated measures data, where data is the log-transformed counts obtained from voom.

Usage

glsSymm(d)

Arguments

d

a data frame containing several columns. The first 4 columns are y: a vector of log-counts (obtained by voom), Subject: a vector of subject/experimental units where repeated measures are obtained (can be either numeric or factor), Time: a vector of time points (continuous, since we fit corSymm), w: weights to put in gls model, this is the inverse of weights obtained by voom The other columns are exactly the same as design matrix.

Value

Output is a vector including the following components

aic	AIC of the fitted model.
s2	estimate of error variance.
rho	correlation parameter in corSymm correlation matrix.
fixed	fixed effects (estimates of regression parameters).
varbeta	the estimates of variance of fixed effects, just include lower part and diagonal part of the variance-covariance matrix.

Examples

data(res)
data(design)
data(covset)
d <- data.frame(cbind(y = res$ori.res$v$E[1,] ,Subject = covset$ear,
Time = as.integer(as.factor(covset$time)), w = 1/res$ori.res$v$weights[1,], design))
glsout <- rmRNAseq:::glsSymm(d)
glsout

ntyet/rmRNAseq documentation built on July 7, 2023, 11:10 a.m.