ols.eblup.trim: The main function for the FastMix pipeline.

In terrysun0302/FastMix: An OLS-based approach for fitting LMER with applications to gene expression deconvolution analysis

Description

A new analytic pipeline, dubbed as FastMix, that combines the deconvolution step with the downstream analyses based on linear mixed effects regression (LMER) model and a moment-matching algorithm.

Usage

ols.eblup.trim(Des, Y, random = "all", independent = F, trim = 0.5, robust = "FastMix", test = 1, trim.fix = TRUE, min.cond.num=1e-6, bias = 2)

Arguments

Des	the design matrix ordered by gene subject by subject. First column should be identification variable, e.g., ID or subject, and the rest columns are covariates.
Y	vectorized gene expression data.
random	'random' is an index vector that specifies which variable(s) requires random effects – by default, all covariates are paired with a random effect.
independent	specify the correlation structure among random effects. If TRUE, random effects are assumed to be independent.
trim	the trimming percentage when accounting for outliers. Default valie is 0.5 (50%).
robust	Specifies whether robust covariance estimation is implemented and which method to use: "FALSE" for non-robust estimation; "mcd" for the MCD algorithm of Rousseeuw and Van Driessen; "weighted" for the Reweighted MCD; "donostah" for the Donoho-Stahel projection based estimator; "pairwiseQC" for the orthogonalized quadrant correlation pairwise estimator. All these algorithms come from the R package 'robust'. "FastMix" is the proposed trimming method.
test	the test method for DEGs. "1" is Gaussian mixture model, "2" is Anderson-darling normal test. Default value is "1".
trim.fix	Whether only consider trimmed subjects in fix effect estiamtion. The default value is FALSE.
min.cond.num	Matrix invertion is used many times in our method. An ill-posed matrix inverse can have detrimental effects in downstream analysis. min.cond.num is a threshold of the minimum condition number of matrix to be inverted. If cond(A) is less than this value (default: 1e-6), a robust matrix inverse is used instead.
bias	The method for bias-correction step. The default value is 2.

Value

fixed.results	the estimated fix effects and their p-values. They are overall effects shared by all genes.
beta.mat	individual coefficient estimation.
Yhat	fitted response.
sigma.beta	the covariance estimation of fixed effect.
VC	variance component estimation. The first column is the one for common random error. The second column is the one for random effects.
cov	???
var.epsilon	the variance of the i.i.d. noise.
var.eblup.mean	the average of the variance of gamma.hat based on the EBLUP estimator. Note that in general, each gamma.hat.i has its own covariance matrix; so var.eblup.mean is provided only as a rough debugging tool.
eta	the chi sqiare type statsitics used for p-value calculation.
re.pvalue	the overall p-value for detecting outliers in random effects.
re.ind.pvalue	the individual p-value for outlier detection for each random effect.
out_idx	he potential covariates with outliers when robust = "FastMix. It is NULL when robust != "FastMix"

Author(s)

Hao Sun

terrysun0302/FastMix documentation built on Nov. 14, 2019, 4:54 a.m.