zinc <- read.table('c:/philip/stat 505/zinc.txt', as.is=T,header=T)
# read the data file, first line gives variable names
#  I use as.is=T to retain data in original format 
#   (otherwise R converts non-numeric to factor variables)

# substitution methods

mean(ifelse(zinc$zinc==7,0,zinc$zinc))
[1] 6.098857

mean(ifelse(zinc$zinc==7,3.5,zinc$zinc))
[1] 7.798857

mean(ifelse(zinc$zinc==7,7,zinc$zinc))
[1] 9.498857



# use of NADA functions:

library(NADA)

zinc.ros <- ros(zinc$zinc,zinc$cens,forwardT=NULL, reverseT=NULL)
# arguments are:
#    x values, 
#    censoring flag (T = censored)
#    forward transformation, default = log
#    backward transformation, default = exp

# estimates from: 
mean(zinc.ros)
sd(zinc.ros)

# probability plot:
plot(zinc.ros)

# details of the underlying regression:
summary(zinc.ros)

# mle methods
zinc.mle <- cenmle(zinc$zinc,zinc$cens, dist='gaussian', conf.int=0.95)
# arguments are observations, censoring flag (T=censored)
#  distribution (default is lognormal)
#  confidence interval coverage (default 0.95)
# note: with gaussian, considers <dl as interval censored (0,dl)

mean(zinc.mle)
sd(zinc.mle)
plot(zinc.mle)
#  produces a quantile-quantile plot 
summary(zinc.mle)
#  all the details on the estimation


# Kaplan-Meier methods
zinc.km <- cenfit(Cen(zinc$zinc,zinc$cens))
#  Cen combines values and censoring flag
#   both are argument to cenfit
# note that values < smallest dl are replaced with that dl

mean(zinc.km)
    rmean se(rmean) 
9.4988571 0.4535363 

sd(zinc.km)
plot(zinc.km)