Multi-environment trial with structured missing values
Grain yield was measured on 5 genotypes in 26 environments. Missing values were non-random, but structured.
envenvironment, 26 levels
gengenotype factor, 5 levels
yieldyield
Used with permission of Jean-Baptists Denis.
Denis, J. B. and C P Baril, 1992, Sophisticated models with numerous missing values: The multiplicative interaction model as an example. Biul. Oceny Odmian, 24–25, 7–31.
H P Piepho, (1999) Stability analysis using the SAS system, Agron Journal, 91, 154–160. https://doi.og/10.2134/agronj1999.00021962009100010024x
## Not run:
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
## End(Not run)Please choose more modern alternatives, such as Google Chrome or Mozilla Firefox.