flm {collapse}

Fast (Weighted) Linear Model Fitting

Description

flm is a fast linear model command that (by default) only returns a coefficient matrix. 6 different efficient fitting methods are implemented: 4 using base R linear algebra, and 2 utilizing the RcppArmadillo and RcppEigen packages. The function itself only has an overhead of 5-10 microseconds, and is thus well suited as a bootstrap workhorse.

Usage

flm(...)  # Internal method dispatch: default if is.atomic(..1)

## Default S3 method:
flm(y, X, w = NULL, add.icpt = FALSE, return.raw = FALSE,
    method = c("lm", "solve", "qr", "arma", "chol", "eigen"),
    eigen.method = 3L, ...)

## S3 method for class 'formula'
flm(formula, data = NULL, weights = NULL, add.icpt = TRUE, ...)

Arguments

y

a response vector or matrix. Multiple dependent variables are only supported by methods "lm", "solve", "qr" and "chol".

X

a matrix of regressors.

w

a weight vector.

add.icpt

logical. TRUE adds an intercept column named '(Intercept)' to X.

formula

a lm formula, without factors, interaction terms or other operators (:, *, ^, -, etc.), may include regular transformations e.g. log(var), cbind(y1, y2), magrittr::multiply_by(var1, var2), magrittr::raise_to_power(var, 2).

data

a named list or data frame.

weights

a weights vector or expression that results in a vector when evaluated in the data environment.

return.raw

logical. TRUE returns the original output from the different methods. For 'lm', 'arma' and 'eigen', this includes additional statistics such as residuals, fitted values or standard errors. The other methods just return coefficients but in different formats.

method

an integer or character string specifying the method of computation:

Int.		String		Description
1		"lm"		uses .lm.fit.
2		"solve"		solve(crossprod(X), crossprod(X, y)).
3		"qr"		qr.coef(qr(X), y).
4		"arma"		uses RcppArmadillo::fastLmPure.
5		"chol"		chol2inv(chol(crossprod(X))) %*% crossprod(X, y) (quite fast, requires crossprod(X) to be positive definite i.e. problematic if multicollinearity).
6		"eigen"		uses RcppEigen::fastLmPure (very fast but, depending on the method, also unstable if multicollinearity).

eigen.method

integer. Select the method of computation used by RcppEigen::fastLmPure:

Int.		Description
0		column-pivoted QR decomposition.
1		unpivoted QR decomposition.
2		LLT Cholesky.
3		LDLT Cholesky.
4		Jacobi singular value decomposition (SVD).
5		method based on the eigenvalue-eigenvector decomposition of X'X.

See vignette("RcppEigen-Introduction", package = "RcppEigen") for details on these methods and benchmark results. Run source(system.file("examples", "lmBenchmark.R", package = "RcppEigen")) to re-run the benchmark on your machine.

...

further arguments passed to other methods. For the formula method further arguments passed to the default method. Additional arguments can also be passed to the default method e.g. tol = value to set a numerical tolerance for the solution - applicable with methods "lm", "solve" and "qr" (default is 1e-7), or LAPACK = TRUE with method "qr" to use LAPACK routines to for the qr decomposition (typically faster than the LINPACK default).

Value

If return.raw = FALSE, a matrix of coefficients with the rows corresponding to the columns of X, otherwise the raw results from the various methods are returned.

Note

Method "qr" supports sparse matrices, so for an X matrix with many dummy variables consider method "qr" passing as(X, "dgCMatrix") instead of just X.

See Also

fhdwithin/HDW, fFtest, Data Transformations, Collapse Overview

Examples

# Simple usage
coef <- flm(mpg ~ hp + carb, mtcars, w = wt)

# Same thing in programming usage
flm(mtcars$mpg, qM(mtcars[c("hp","carb")]), mtcars$wt, add.icpt = TRUE)

# Check this is correct
lmcoef <- coef(lm(mpg ~ hp + carb, weights = wt, mtcars))
all.equal(drop(coef), lmcoef)

# Multi-dependent variable (only some methods)
flm(cbind(mpg, qsec) ~ hp + carb, mtcars, w = wt)

# Returning raw results from solver: different for different methods
flm(mpg ~ hp + carb, mtcars, return.raw = TRUE)
flm(mpg ~ hp + carb, mtcars, method = "qr", return.raw = TRUE)
 
# Test that all methods give the same result
all_obj_equal(lapply(1:6, function(i)
  flm(mpg ~ hp + carb, mtcars, w = wt, method = i)))