theobald.covariate {agridat}R Documentation

Multi-environment trial of corn silage, Year * Loc * Variety with covariate

Description

Corn silage yields for maize in 5 years at 7 districts for 10 hybrids.

Format

A data frame with 256 observations on the following 5 variables.

year

year, 1990-1994

env

environment/district, 1-7

gen

genotype, 1-10

yield

dry-matter silage yield for corn

chu

corn heat units, thousand degrees Celsius

Details

The trials were carried out in seven districts in the maritime provinces of Eastern Canada. Different fields were used in successive years. The covariate CHU (Corn Heat Units) is the accumulated average daily temperatures (thousands of degrees Celsius) during the growing season at each location.

Used with permission of Chris Theobald.

Source

Chris M. Theobald and Mike Talbot and Fabian Nabugoomu, 2002. A Bayesian Approach to Regional and Local-Area Prediction From Crop Variety Trials, Journ Agric Biol Env Sciences, 7, 403–419. http://doi.org/10.1198/108571102230

Examples


data(theobald.covariate)
dat <- theobald.covariate
require(lattice)
xyplot(yield ~ chu|gen, dat, type=c('p','smooth'),
       xlab =  "chu = corn heat units",
       main="theobald.covariate - yield vs heat")

## Not run: 
  # REML estimates (Means) in table 3 of Theobald 2002
  require(lme4)
  dat <- transform(dat, year=factor(year))
  m0 <- lmer(yield ~ -1 + gen + (1|year/env) + (1|gen:year), data=dat)
  round(fixef(m0),2)

## End(Not run)

## Not run: 

# Use JAGS to fit Theobald (2002) model 3.2 with 'Expert' prior

require(reshape2)
ymat <- acast(dat, year+env~gen, value.var='yield')
chu <- acast(dat, year+env~., mean, value.var='chu', na.rm=TRUE)
chu <- as.vector(chu - mean(chu))  # Center the covariate
dat$yr <- as.numeric(dat$year)
yridx <- as.vector(acast(dat, year+env~., mean, value.var='yr', na.rm=TRUE))
dat$loc <- as.numeric(dat$env)
locidx <- acast(dat, year+env~., mean, value.var='loc', na.rm=TRUE)
locidx <- as.vector(locidx)

jdat <- list(nVar = 10, nYear = 5, nLoc = 7, nYL = 29, yield = ymat,
            chu = chu, year = yridx, loc = locidx)

require(rjags)
m1 <- jags.model(file=system.file(package="agridat", "files/theobald.covariate.jag"),
  data=jdat, n.chains=2)

# Table 3, Variety deviations from means (Expert prior)
c1 <- coda.samples(m1, variable.names=(c('alpha')),
                   n.iter=10000, thin=10)
s1 <- summary(c1)
effs <- s1$statistics[,'Mean']
rev(sort(round(effs - mean(effs), 2))) # Perfect match (different order?)

## End(Not run)


[Package agridat version 1.16 Index]