agridat (version 1.16)

hughes.grapes: Disease incidence on grape leaves in RCB experiment

Description

Disease incidence on grape leaves in RCB experiment with 6 different treatments.

Format

A data frame with 270 observations on the following 6 variables.

block

Block factor, 1-3

trt

Treatment factor, 1-6

vine

Vine factor, 1-3

shoot

Shoot factor, 1-5

diseased

Number of diseased leaves per shoot

total

Number of total leaves per shoot

Details

These data come from a study of downy mildew on grapes. The experiment was conducted at Wooster, Ohio, on the experimental farm of the Ohio Agricultural Research and Development Center, Ohio State University.

There were 3 blocks with 6 treatments. Treatment 1 is the unsprayed control. On 30 Sep 1990, disease incidence was measured. For each plot, 5 randomly chosen shoots on each of the 3 vines were observed. The canopy was closed and shoots could be intertwined. On each shoot, the total number of leaves and the number of infected leaves were recorded.

References

Hans-Pieter Piepho. 1999. Analysing disease incidence data from designed experiments by generalized linear mixed models. Plant Pathology, 48, 668--684. http://doi.org/10.1046/j.1365-3059.1999.00383.x

Examples

# NOT RUN {
data(hughes.grapes)
dat <- hughes.grapes

dat <- transform(dat, rate = diseased/total, plot=trt:block)

# Trt 1 has higher rate, more variable, Trt 3 lower rate, less variable
require(lattice)
foo <- bwplot(rate ~ vine|block*trt, dat, main="hughes.grapes",
              xlab="vine")
if(require(latticeExtra)){
  useOuterStrips(foo)
}

# Table 1 of Piepho 1999
tapply(dat$rate, dat$trt, mean) # trt 1 does not match Piepho
tapply(dat$rate, dat$trt, max)

# ----------------------------------------------------------------------------

# }
# NOT RUN {
  # Piepho model 3.  Binomial data.  May not be exactly the same model
  
  # Use the binomial count data with lme4
  require(lme4)
  m1 <- glmer(cbind(diseased, total-diseased) ~ trt + block + (1|plot/vine),
              data=dat, family=binomial)
  m1

  # Switch from binomial counts to bernoulli data
  require(aod)
  bdat <- splitbin(cbind(diseased, total-diseased) ~ block+trt+plot+vine+shoot,
                   data=dat)$tab
  names(bdat)[2] <- 'y'
  
  # Using lme4
  m2 <- glmer(y ~ trt + block + (1|plot/vine), data=bdat, family=binomial)
  m2
  
  # Now using MASS:::glmmPQL
  require(MASS)
  m3 <- glmmPQL(y ~ trt + block, data=bdat,
                random=~1|plot/vine, family=binomial)
  m3
# }
# NOT RUN {
# }