cullis.earlygen: Early generation variety trial in wheat

Description Format Details Source References Examples

Description

Early generation variety trial in wheat

Format

A data frame with 670 observations on the following 5 variables.

gen

genotype factor

row

row

col

column

entry

entry (genotype) number

yield

yield of each plot, kg/ha

weed

weed score

Details

The data are from an unreplicated field experiment conducted at Tullibigeal, New South Wales, Australia in 1987-88. In each row, every 6th plot was the variety 'Kite'. Six other standard varieties were randomly interspersed over the trial. Each plot was 15m x 1.8m, "oriented with the longest side with rows".

The 'weed' variable is a visual score on a 0 to 10 scale, 0 = no weeds, 10 = 100 percent weeds.

The replicated check variety was numbered 526. A further 6 replicated commercially available varieties (numbered 527 to 532) were also randomly assigned to plots with between 3 to 5 plots of each. The aim of these trials is to identify and retain the top, say 20 percent of lines for further testing. Cullis et al. (1989) presented an analysis of early generation variety trials that included a one-dimensional spatial analysis. Below, a two-dimensional spatial analysis is presented.

Note: The 'row' and 'col' variables are as in the VSN link below (switched compared to the paper by Cullis et al.)

Field width: 10 rows * 15 m = 150 m

Field length: 67 plots * 1.8 m = 121 m

The orientation is not certain, but the alternative orientation would have a field roughly 20m x 1000m, which seems unlikely.

Source

Brian R. Cullis, Warwick J. Lill, John A. Fisher, Barbara J. Read and Alan C. Gleeson (1989). A New Procedure for the Analysis of Early Generation Variety Trials. Journal of the Royal Statistical Society. Series C (Applied Statistics), 38, 361-375. http://doi.org/10.2307/2348066

References

Unreplicated early generation variety trial in Wheat. http://www.vsni.co.uk/software/asreml/htmlhelp/asreml/xwheat.htm

Examples

  1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
data(cullis.earlygen)
dat <- cullis.earlygen

# Show field layout of checks.  Cullis Table 1.
dat$check <- ifelse(dat$entry < 8, dat$entry, NA)
if(require(desplot)){
  desplot(yield ~ col*row, dat,
          col="check", cex=0.5, flip=TRUE, aspect=121/150, # true aspect
          main="cullis.earlygen (yield)")

  grays <- colorRampPalette(c("white","#252525"))
  desplot(weed ~ col*row, dat,
          at=0:6-0.5, col.regions=grays(7)[-1],
          flip=TRUE, aspect=121/150, # true aspect
          main="cullis.earlygen (weed)")
}

require(lattice)
bwplot(yield ~ as.character(weed), dat,
       horizontal=FALSE,
       xlab="Weed score", main="cullis.earlygen")

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

## Not run: 
  # asreml3
  require(asreml)
  # Start with the standard AR1xAR1 analysis
  dat <- transform(dat, xf=factor(col), yf=factor(row))
  dat <- dat[order(dat$xf, dat$yf),]
  m2 <- asreml(yield ~ weed, data=dat, random= ~gen,
               rcov = ~ ar1(xf):ar1(yf))
  
  # Variogram suggests a polynomial trend
  m3 <- update(m2, fixed= yield~weed+pol(col,-1))
  
  # Now add a nugget variance
  m4 <- update(m3, random= ~ gen + units)
  
  require(lucid)
  vc(m4)
  ##           effect component std.error z.ratio constr
  ##      gen!gen.var  73770    10420         7.1    pos
  ##  units!units.var  30440     8074         3.8    pos
  ##       R!variance  54720    10630         5.1    pos
  ##         R!xf.cor      0.38     0.115     3.3  uncon
  ##         R!yf.cor      0.84     0.045    19    uncon
  
  # Predictions from models m3 and m4 are non-estimable.  Why?
  # Use model m2 for predictions
  predict(m2)$pred
  ##         gen predicted.value standard.error est.status
  ## 1    Banks         2723.534       93.14633  Estimable
  ## 2    Eno008        2981.057      162.85053  Estimable
  ## 3    Eno009        2978.009      161.56930  Estimable
  ## 4    Eno010        2821.399      153.96697  Estimable
  ## 5    Eno011        2991.610      161.53308  Estimable
  ## 6    Eno012        2771.148      162.21897  Estimable

  # Compare AR1 with Moving Grid
  require(mvngGrAd)
  shape <- list(c(1),
                c(1),
                c(1:4),
                c(1:4))
  # sketchGrid(10,10,20,20,shapeCross=shape, layers=1, excludeCenter=TRUE)
  m5 <- movingGrid(rows=dat$row, columns=dat$col, obs=dat$yield,
                   shapeCross=shape, layers=NULL)
  dat$mg <- fitted(m5)
  dat$ar1 <- fitted(m2)
  head(dat[ , c('yield','ar1','mg')])
  ##    yield      ar1       mg
  ## 1   2652 2467.980 2531.998
  ## 11  3394 3071.681 3052.160
  ## 21  3148 2826.188 2807.031
  ## 31  3426 3026.985 3183.649
  ## 41  3555 3070.102 3195.910
  ## 51  3453 3006.352 3510.511
  pairs(dat[ , c('yield','ar1','mg')])
  

## End(Not run)

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

## Not run: 
  ## require(asreml4)
  ## # Start with the standard AR1xAR1 analysis
  ## dat <- transform(dat, xf=factor(col), yf=factor(row))
  ## dat <- dat[order(dat$xf, dat$yf),]
  ## m2 <- asreml(yield ~ weed, data=dat, random= ~gen,
  ##              resid = ~ ar1(xf):ar1(yf))
  
  ## # Variogram suggests a polynomial trend
  ## m3 <- update(m2, fixed= yield~weed+pol(col,-1))
  
  ## # Now add a nugget variance
  ## m4 <- update(m3, random= ~ gen + units)
  
  ## require(lucid)
  ## vc(m4)
  ## ##       effect component std.error z.ratio bound 
  ## ##          gen  73780    10420         7.1     P 0  
  ## ##        units  30440     8073         3.8     P 0.1
  ## ##     xf:yf(R)  54730    10630         5.1     P 0  
  ## ## xf:yf!xf!cor      0.38     0.115     3.3     U 0  
  ## ## xf:yf!yf!cor      0.84     0.045    19       U 0  
  
  ## # Predictions from models m3 and m4 are non-estimable.  Why?
  ## # Use model m2 for predictions
  ## predict(m2, classify="gen")$pvals
  ## ##         gen predicted.value std.error    status
  ## ## 1     Banks        2723.534  93.14719 Estimable
  ## ## 2    Eno008        2981.056 162.85241 Estimable
  ## ## 3    Eno009        2978.008 161.57129 Estimable
  ## ## 4    Eno010        2821.399 153.96943 Estimable
  ## ## 5    Eno011        2991.612 161.53507 Estimable
  

  ## # Compare AR1 with Moving Grid
  ## require(mvngGrAd)
  ## shape <- list(c(1),
  ##               c(1),
  ##               c(1:4),
  ##               c(1:4))
  ## # sketchGrid(10,10,20,20,shapeCross=shape, layers=1, excludeCenter=TRUE)
  ## m5 <- movingGrid(rows=dat$row, columns=dat$col, obs=dat$yield,
  ##                  shapeCross=shape, layers=NULL)
  ## dat$mg <- fitted(m5)
  ## dat$ar1 <- fitted(m2)
  ## head(dat[ , c('yield','ar1','mg')])
  ## ##    yield      ar1       mg
  ## ## 1   2652 2467.980 2531.998
  ## ## 11  3394 3071.681 3052.160
  ## ## 21  3148 2826.188 2807.031
  ## ## 31  3426 3026.985 3183.649
  ## ## 41  3555 3070.102 3195.910
  ## ## 51  3453 3006.352 3510.511
  ## pairs(dat[ , c('yield','ar1','mg')])


## End(Not run)

agridat documentation built on May 2, 2019, 4:01 p.m.