irwls.diag<-function(data,glmobj){
#
# Function to create IRWLS-based diagnostics corresponding to glm object.
#
# Input:
#        data    - data frame containing data used to create glm object
#        glmobj  - glm object defining model
#
# Output:
#        stdres  - standardized residuals
#        hat     - leverage values
#        cook    - Cook's distances
#        phi     - index of overdispersion. If overdispersion-corrected version
#                  of the model is desired, standardized residuals must be divided
#                  by the square root of phi
#        summary - summary of overdispersion-corrected version of the model
#
  attach(data)
  fit<-predict.gam(glmobj,type="response")
  u<-log(fit)+(glmobj$y-fit)/fit
  ll<-lm(update(formula(glmobj),u~.),weights=fit,qr=T)
  phi<-deviance(ll)/ll$df.residual
  hat<-hat(ll$qr)
  stdres<-residuals(ll,type="pearson")/sqrt(1-hat)
  cook<-((stdres^2)*hat)/(length(ll$coeff)*(1-hat)*phi)
  summary<-summary(ll)
  detach()
  return(list(stdres=stdres,hat=hat,cook=cook,phi=phi,summary=summary))}