[R] glm() scale parameters and predicted Values

Peter Maclean pmaclean2011 at yahoo.com
Thu Jul 14 06:15:59 CEST 2011

In glm() you can use the summary() function to recover the shape parameter (the reciprocal of the dispersion parameter). How do you recover the scale parameter? Also, in the given example, how I estimate and save the geometric mean of the predicted values? For a simple model you can use fitted() or predicted() functions. I will appreciate any help. 
#Call required R packages
#Grouped vector
n <- c(1:10)
yr <-c(1:10)
ny <- list(yr=yr,n=n)
ny <- expand.grid(ny) 
y = rgamma(100, shape=1.5, rate = 1, scale = 2)
Gdata <- cbind(ny,y)
Gdata2<- Gdata
Gdata$x1 <- cos((3.14*yr)/365.25) 
Gdata$x2 <- sin((3.14*yr)/365.25) 
#Fitting Generalized Linear Models 
Gdata <- split(Gdata,Gdata$n)
FGLM <- lapply(Gdata, function(x){
              m <- as.numeric(x$y)
              x1 <- m <- as.numeric(x$x1)
              x2 <- m <- as.numeric(x$x2)
              summary(glm(m~1+x1+x2, family=Gamma),dispersion=NULL) 

#Save the results of the estimated parameters
str(FGLM,no.list = TRUE)
SFGLMC<- ldply(FGLM, function(x) x$coefficients)
SFGLMD<- ldply(FGLM, function(x) x$dispersion)

More information about the R-help mailing list