wls_est: Blockwise WLS estimation
In JanMarvin/nlsur: Estimating Nonlinear Least Squares for Equation Systems

wls_est

R Documentation

Blockwise WLS estimation

Description

Blockwise WLS estimation. Usually for (X'X)^{-1} W^{-1} X'Y X and Y X, W and Y are of similar dimensions. In nlsur W is a cov-matrix of size k \times k and usually way smaller than X. To avoid blowing all matrices up for the estimation, a blockwise approach is used. X is shrunken to match size k. W is D'D so XDX is calculated. XDy is only calculated if wanted for a full WLS. For the cov-matrix only XDX is required.

Usage

wls_est(x, r, qS, w, sizetheta, fullreg, tol)

Arguments

`x`	matrix of derivatives
`r`	residual matrix
`qS`	weighting matrix of sizetheta x sizetheta
`w`	vector of weights
`sizetheta`	integer defining the amount of coefficients
`fullreg`	bool defining if WLS or Cov is calculated
`tol`	tolerance used for qr()