dot-nlsur: Non-Linear Seemingly Unrelated Regression
In JanMarvin/nlsur: Estimating Nonlinear Least Squares for Equation Systems

.nlsur

R Documentation

Non-Linear Seemingly Unrelated Regression

Description

.nlsur() is a function for estimation of a non-linear seemingly unrelated regression model in R.

Usage

.nlsur(
  eqns,
  data,
  startvalues,
  S = NULL,
  robust = robust,
  nls = FALSE,
  fgnls = FALSE,
  ifgnls = FALSE,
  qrsolve = FALSE,
  MASS = FALSE,
  trace = FALSE,
  eps = eps,
  tau = tau,
  maxiter = maxiter,
  tol = tol,
  initial = initial
)

Arguments

`eqns`	is can be a single equation or a equation system. If eqns is a single equation it will internally be converted to a list. Estimation of a single equation might as well be done using `nls()`.
`data`	is the data set on which the equation is applied. This can be of every type `eval()` can handle.
`startvalues`	is a vector of initial start values. For
`S`	is a weighing matrix used for estimation in Feasible Generalized Non-Linear Least Squares (FGNLS) and Iterative FGNLS. For `nlsur()` this is assumed to be the identity matrix. Hence, it is not included. If included S is expected to be a matrix.
`robust`	should a robust standard error be calculated
`nls`	is a logical and default if estimation is done for NLSUR or NLS.
`fgnls`	is a logical and must be set, if estimation is done for FGNLS. This is called in a function called `fgnls()` and should not be set by the user.
`ifgnls`	is a logical and must be set, if estimation is done for ifgnls. This is called in a function called `nlsur()` and should not be set by the user.
`qrsolve`	is a logical, if TRUE `qr.coef(qr(x), r)` is called which should be the most robust way for estimation of nls. For this all equations will be rbinded, which might lead to memory bottlenecks.
`MASS`	is a logical, if TRUE `lm_gls()` is called for estimation of a linear regression with a weighting matrix.
`trace`	is a logical. If TRUE the current iterations SSR is called.
`eps`	the epsilon used for convergence in nlsur(). Default is 1e-5.
`tau`	is another convergence variable. Default is 1e-3.
`maxiter`	maximum number of iterations
`tol`	qr.solves tolerance for detecting linear dependencies.
`initial`	logical if initial calculation set. used to avoid calculation of svd numerous times

Details

nlsur is a function for estimation of a non-linear least squares (NLS). In addition to nls() it is capable of estimation of system of equations. This estimation is done in a non-linear seemingly unrelated regression approach.