lol (version 1.20.0)

lmMatrixFit: Multiple lm fit for penalized regressions

Description

Refit the regressions given matrices of responses, predictors, and the coefficients/interactions matrix. This is typically used after the lasso, since the coefficients were shrinked.

Usage

lmMatrixFit(y, x = NULL, mat, th = NULL)

Arguments

y

Input response matrix, typically expression data with genes/variables in columns and samples/measurements in rows. Or when input x is NULL, y should be an object of two lists: y: expression data and x: copy number data

x

Input predictor matrix, typically copy number data, genes/predictors in columns and samples/measurements in rows. Can be NULL

mat

Coefficient matrix, number of columns is the number of predictors (y) and number of rows is the number of responses (x)

th

The threshold to use in order to determine which coefficients are non-zero, so the corresponding predictors are used

Value

coefMat

A coefficient matrix, rows are responses and columns are predictors

resMat

A residual matrix, each row is the residuals of a response.

pvalMat

Matrix of p-values for each coefficients

See Also

lm, matrixLasso

Examples

data(chin07)
data <- list(y=t(chin07$ge), x=t(chin07$cn))
res <- matrixLasso(data, method='cv', nFold=5)
res
res.lm <- lmMatrixFit(y=data, mat=abs(res$coefMat), th=0.01)
res.lm