| Library: | xplore |
| See also: | linreg glmest glmstat |
| Quantlet: | gls | |
| Description: | Computes the Generalized Least Squares estimate for the coefficients of a linear model when the errors have a positive definite covariance matrix om. |
| Usage: | b = gls (x, y{, om}) | |
| Input: | ||
| x | n x p x d1 x ... x dn array, explanatory variables | |
| y | n x 1 x d1 x ... x dn array, dependent variable | |
| om | optional, n x n x d1 x ... x dn array, covariance matrix | |
| Output: | ||
| b | p x 1 x d1 x ... x dn array, GLS estimates | |
library("plot") ; load libraries
library("stats")
x=read("nicfoo") ; read data
xd = matrix(rows(x)) ~ x[,1] ~ x[,1]^2
yd = x[,2]
b = gls(xd, yd) ; OLS
eps = yd - xd * b ; calculate residuals
sh = sknn(x[,1],eps^2,5) ; smooth the variance
om = diag(sh) ; create covariance matrix
bg = gls(xd, yd, om) ; GLS with om
b ~ bg ; OLS vs. GLS
Contents of _tmp [1,] 0.11264 0.11975 [2,] 1.2295 1.2065 [3,] -0.25594 -0.24865