agridat (version 1.23)

denis.missing: Multi-environment trial with structured missing values

Description

Grain yield was measured on 5 genotypes in 26 environments. Missing values were non-random, but structured.

Format

env

environment, 26 levels

gen

genotype factor, 5 levels

yield

yield

Used with permission of Jean-Baptists Denis.

References

H P Piepho, (1999) Stability analysis using the SAS system, Agron Journal, 91, 154--160. https://doi.og/10.2134/agronj1999.00021962009100010024x

Examples

if (FALSE) {

library(agridat)
data(denis.missing)
dat <- denis.missing

# view missingness structure
libs(reshape2)
acast(dat, env~gen, value.var='yield')


libs(lattice)
redblue <- colorRampPalette(c("firebrick", "lightgray", "#375997"))
levelplot(yield ~ gen*env, data=dat,
          col.regions=redblue,
          main="denis.missing - incidence heatmap")

# stability variance (Table 3 in Piepho)
libs(nlme)
m1 <- lme(yield ~ -1 + gen, data=dat, random= ~ 1|env,
          weights = varIdent(form= ~ 1|gen),
          na.action=na.omit)
svar <- m1$sigma^2 * c(1, coef(m1$modelStruct$varStruct, unc = FALSE))^2
round(svar, 2)
##          G5    G3    G1    G2
## 39.25 22.95 54.36 12.17 23.77

}