Multi-environment trial with structured missing values

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.

Source

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.

References

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

Examples

# \dontrun{

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

# view missingness structure
libs(reshape2)
acast(dat, env~gen, value.var='yield')
#>     G1 G2 G3 G4 G5
#> E01 NA NA NA 66 49
#> E02 NA NA NA 59 40
#> E03 NA NA NA 67 57
#> E04 NA NA NA 55 53
#> E05 NA NA 75 85 74
#> E06 NA NA 72 77 71
#> E07 NA NA 55 51 46
#> E08 NA NA 52 49 44
#> E09 NA NA 79 80 62
#> E10 63 64 72 65 51
#> E11 72 82 83 88 72
#> E12 69 76 78 86 70
#> E13 55 53 40 63 60
#> E14 52 59 54 53 51
#> E15 67 62 65 68 65
#> E16 63 74 73 63 59
#> E17 55 54 NA 52 51
#> E18 66 66 NA NA 66
#> E19 57 54 NA NA NA
#> E20 69 64 NA NA NA
#> E21 73 69 NA NA NA
#> E22 68 66 NA NA NA
#> E23 64 79 NA NA NA
#> E24 67 66 NA NA NA
#> E25 70 66 NA NA NA
#> E26 70 65 NA 60 NA


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 
##          G5    G3    G1    G2
## 39.25 22.95 54.36 12.17 23.77

# }