lmspar: fit a linear model with spatial CAR and SAR covariances using...

In jayverhoef/SpAReco: Paper on Spatial Autoregressive Models for Ecological Inference

Description

fit a linear model with spatial CAR and SAR covariances using maximum likelihood

Usage

lmspar(formula, data, indComp = FALSE, Nmat = NULL, distMat = NULL,
  crsStk = NULL, indSamp, model = NULL, rowStand = TRUE,
  rhoBound = c(-1, 1))

Arguments

formula	an R formula
data	a data.frame containing the data
indComp	should an independent component be added to the model
Nmat	neighborhood matrix of binary values, with a 1 indicating a neighbor, 0 otherwise
distMat	Euclidean distance matrix among polygon centroids
crsStk	indicator matrix of whether locations cross stock boundaries
indSamp	indicator vector for wheter location was sampled. Zero indicates missing value
model	either 'CAR' or 'SAR'. Only used if one of Nmat, distMat, or crsStk are not NULL.
rowStand	should row-standardization be used on CAR or SAR models? Default is TRUE.

Value

a list with parmEst as the estimated covariance parameters (with overall variance profiled out), sigma2 as the estimate of the overall variance parameter, m2LL as the minimized negative 2 times the log-likelihood, bHat as the fitted fixed effect parameters, covb as the estimated covariance matrix of the fitted fixed effect parameters, Vi is the inverse of the covariance matrix, Vi.oo as the inverse of the covariance matrix of the observed locations only, X returns the design matrix, and y returns the data vector.

Author(s)

Jay Ver Hoef

jayverhoef/SpAReco documentation built on May 18, 2019, 5:58 p.m.