Ornstein1 <- read.table("c:/temp/Ornstein.txt",header=TRUE)

Ornstein1$assets2  <- Ornstein1$assets/1000


Ornstein1.glm1 <- glm(interlocks ~ as.factor(sector) +as.factor(nation) +assets2,
                        data=Ornstein1,family=poisson())

summary(Ornstein1.glm1)

par(mfrow=c(2,2)) 
plot(Ornstein1.glm1)
par(mfrow=c(1,1)) 


 # These three reduced models give the residual deviances in the Table on page 388

summary(glm(interlocks ~ as.factor(sector) +as.factor(nation) ,
                        data=Ornstein1,family=poisson()) )


summary(glm(interlocks ~ as.factor(sector) +assets2,
                        data=Ornstein1,family=poisson()) )


summary(glm(interlocks ~ as.factor(nation) +assets2,
                        data=Ornstein1,family=poisson()) )



 # the effects package can produce effects plots like those on page 390

library(effects)

 # the as.factor function cannot be used in the glm formula with the effects
 # functions, so here they are defined as separate variables

Ornstein1$fnation <-  as.factor(Ornstein1$nation)

Ornstein1$fsector <-  as.factor(Ornstein1$sector)

Ornstein1.glm1a <- glm(interlocks ~ fsector +fnation +assets2,
                        data=Ornstein1,family=poisson())


plot(effect("fsector", Ornstein1.glm1a))

plot(effect("fnation", Ornstein1.glm1a))

plot(effect("assets2", Ornstein1.glm1a))