So stay with errorsarlm() for error SAR models, or spautolm() for a wider
selection of error models.
=>
?spautolm, which shows
Spatial conditional and simultaneous autoregression model estimation
Description:
Function taking family and weights arguments for spatial
autoregression model estimation by Maximum Likelihood, using full
matrix methods, not suited to large data sets with thousands of
obse'interval=' argument should be set.
both SAR and CAR were tried but run very well with 4000 data sets here.
In the results, CAR seems to be more suitable than SAR based on
log-likelihood and AIC.
Elaine
Results
> summary(mignb8.sar)
Call: spautolm(formula = datam.std$SummerM_ratio ~ datam.std$coast +
datam.std$topo_mean + datam.std$prec_mean, data = datam.std, listw =
nb8.w, na.action = na.omit,
family = "SAR", method = "Matrix", verbose = TRUE, zero.policy = TRUE)
Residuals:
Min 1Q Median 3Q Max
-0.4775092 -0.0438067 -0.0045993 0.0405448 0.4616156
Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) 3.8716032 1.2590572 3.0750 0.002105
datam.std$coast 0.0149182 0.0052961 2.8169 0.004850
datam.std$topo_mean 0.0142579 0.0024689 5.7751 7.690e-09
datam.std$prec_mean -0.0427762 0.0033422 -12.7988 < 2.2e-16
Lambda: 0.999 LR test value: 13678 p-value: < 2.22e-16
Log likelihood: 4909.326
ML residual variance (sigma squared): 0.007599, (sigma: 0.087172)
Number of observations: 4873
Number of parameters estimated: 8
AIC: -9802.7
> summary(mignb8.car)
Call: spautolm(formula = datam.std$SummerM_ratio ~ datam.std$coast +
datam.std$topo_mean + datam.std$prec_mean, data = datam.std, listw =
nb8.w, na.action = na.omit,
family = "CAR", method = "Matrix", verbose = TRUE, zero.policy = TRUE)
Residuals:
Min 1Q Median 3Q Max
-0.487501 -0.086579 0.011347 0.116020 0.970053
Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) 1.7232148 0.0516192 33.383 < 2.2e-16
datam.std$coast 0.3527424 0.0050680 69.603 < 2.2e-16
datam.std$topo_mean 0.0563818 0.0029408 19.172 < 2.2e-16
datam.std$prec_mean -0.0722950 0.0040828 -17.707 < 2.2e-16
Lambda: 0.999 LR test value: 11133 p-value: < 2.22e-16
Log likelihood: 3636.802
ML residual variance (sigma squared): 0.012985, (sigma: 0.11395)
Number of observations: 4873
Number of parameters estimated: 8
AIC: -7257.6
code
rm(list=ls())
datam <-read.csv("c:/migration/Mig_ratio_20100808.csv",header=T,
row.names=1)
dim(datam)
library(ncf)
# Define coordinates, neighbours, and spatial weights
coords<-cbind(datam$lon,datam$lat)
coords<-as.matrix(coords)
# Define neighbourhood (here distance 8)
nb8<-dnearneigh(coords,0,8.12)
summary(nb8)
# Spatial weights, illustrated with coding style "W" (row standardized)
nb8.w<-nb2listw(nb8, glist=NULL, style="W")
# std model
datam.sd<-scale(datam)
datam.std<-as.data.frame(datam.sd)
summary (datam.std)
mean(datam.std)
# obtain standard deviation
sd(datam.std)
mig.std
<-lm(datam.std$Summer~datam.std$coast+datam.std$topo_var+datam.std$prec_ran,data=datam.std)
summary(mig.std)
# SAR of spautolm
mignb8.sar <-
spautolm(datam.std$Summer~datam.std$coast+datam.std$topo_var+datam.std$prec_ran,data=datam.std,
listw=nb8.w, na.action=na.omit, , zero.policy=TRUE, family="SAR",
method="Matrix", verbose=TRUE)
summary(mignb8.sar)
# CAR of spautolm
mignb8.car <-
spautolm(datam.std$Summer~datam.std$coast+datam.std$topo_var+datam.std$prec_ran,data=datam.std,
listw=nb8.w, na.action=na.omit, , zero.policy=TRUE, family="CAR",
method="Matrix", verbose=TRUE)
summary(mignb8.car)
>
>
[[alternative HTML version deleted]]
_______________________________________________
R-sig-Geo mailing list
R-sig-Geo@stat.math.ethz.ch
https://stat.ethz.ch/mailman/listinfo/r-sig-geo
