The goal of sppois is to implement the two stage estimator for spatial autoregressive Poisson models proposed by Lambert et al. (2010)
This package does not currently implement the Lambert methods. Contributions welcome.
sppois is not on CRAN. When it is, you can install the released version of sppois from CRAN with:
install.packages("sppois")For now, you can install the development version from GitHub with:
# install.packages("devtools")
devtools::install_github("gregmacfarlane/sppois")The package allows users to estimate a full-information maximum likelihood estimation of spatially-autogressive poisson count models.
library(sppois)
crime_sppois <- sarpoisson(crime_i ~ income + home_value, data = columbus_crime,
listw = columbus_neighbors, method = "fiml")
#> Warning in sarpoisson(crime_i ~ income + home_value, data = columbus_crime, :
#> Matrix listw is near-singular, results may not be reliable.
summary(crime_sppois)
#>
#> Call:
#> sarpoisson(formula = crime_i ~ income + home_value, data = columbus_crime,
#> listw = columbus_neighbors, method = "fiml")
#>
#> Residuals:
#> Min 1Q Median 3Q Max
#> -3.5366373 -0.5430774 0.0953256 0.7218270 2.8718411
#>
#> Coefficients:
#> Estimate Std. Error t value Pr(>|t|)
#> rho 0.35302 0.31453 1.12 0.268
#> (Intercept) 1.65185 0.64807 2.55 0.014 *
#> income -0.04082 0.01914 -2.13 0.038 *
#> home_value -0.00758 0.00515 -1.47 0.148
#> ---
#> Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
#>
#> Log likelihood: -83.16539 for fiml model